---
title: Claude Fable 5 | Runware Docs
url: https://runware.ai/docs/models/anthropic-claude-fable-5
description: Frontier multimodal agentic model for long-horizon coding, research, and complex professional work
---
# Claude Fable 5

![Claude Fable 5](https://assets.runware.ai/covers/anthropic-claude-fable-5.jpg)

Claude Fable 5 is Anthropic's new top-capability generally available Claude model. It is built for long-running coding, agentic execution, multimodal reasoning, research, and high-stakes professional workflows, with stronger long-horizon performance than earlier Opus models, state-of-the-art vision, and a 1M-token context window. Safety classifiers can route some sensitive requests to Claude Opus 4.8 instead.

- **Status**: live
- **ID**: `anthropic-claude-fable-5`
- Try in Playground
- **Context window**: 1M context
- **Max output**: 128K output
- **Creator**: Anthropic
- **Release Date**: June 9, 2026
- **Capabilities**: Text to Text, Image to Text, Checkpoint

## Pricing

~$0.13 typical run

**Rates** (per 1M tokens)

- Input: `$10.00`
- Output: `$50.00`
- Cached input: `$1.00`
- Cache write: `$12.50`

**4 example generations**

- [split-thinking](https://runware.ai/docs/models/anthropic-claude-fable-5/examples#example-commercial-solar-bid-recommendation): `$0.0347` (~11s)
- [text-generation](https://runware.ai/docs/models/anthropic-claude-fable-5/examples#example-library-capital-campaign-case-statement): `$0.119` (~35s)
- [structured-output](https://runware.ai/docs/models/anthropic-claude-fable-5/examples#example-mobile-prenatal-clinic-siting-brief): `$0.14` (~33s)
- [reference-to-text](https://runware.ai/docs/models/anthropic-claude-fable-5/examples#example-freight-damage-claims-assessment): `$0.271` (~1m 25s)

billed per token, so a request costs what the model produced · times vary, measured end to end · [How pricing works](https://runware.ai/docs/models-api/pricing)

## Compatibility & Validation

When `jsonSchema` is provided, `outputFormat` must be `JSON`.

## Request Parameters

**API Options**

Platform-level options for task execution and delivery.

### [taskType](https://runware.ai/docs/models/anthropic-claude-fable-5#request-tasktype)

- **Type**: `string`
- **Required**: true
- **Value**: `textInference`

Identifier for the type of task being performed.

### [taskUUID](https://runware.ai/docs/models/anthropic-claude-fable-5#request-taskuuid)

- **Type**: `string`
- **Required**: true
- **Format**: `UUID v4`

UUID v4 identifier for tracking tasks and matching async responses. Must be unique per task.

### [outputFormat](https://runware.ai/docs/models/anthropic-claude-fable-5#request-outputformat)

- **Type**: `string`
- **Required**: conditional — Required when jsonSchema is provided
- **Default**: `TEXT`

Specifies the file format of the generated output. The available values depend on the task type and the specific model's capabilities.

**Allowed values**: `TEXT` `JSON`

### [webhookURL](https://runware.ai/docs/models/anthropic-claude-fable-5#request-webhookurl)

- **Type**: `string`
- **Format**: `uri`

Webhook URL that receives JSON responses via HTTP POST when generation tasks complete. For batch requests with multiple results, each completed item triggers a separate webhook call as it becomes available.

**Learn more** (1 resource):

- [Webhooks](https://runware.ai/docs/models-api/webhooks) (platform)

### [deliveryMethod](https://runware.ai/docs/models/anthropic-claude-fable-5#request-deliverymethod)

- **Type**: `string`
- **Default**: `sync`

Determines how the API delivers task results.

**Allowed values**:

- `sync` Returns results directly in the API response. Over HTTP, a task still running after 120 seconds continues as an async task and comes back with status 'processing'. Poll for its results using getResponse.
- `async` Returns an immediate acknowledgment with the task UUID. Poll for results using getResponse.
- `stream` Streams results token-by-token as they are generated.

**Learn more** (1 resource):

- [Task Polling](https://runware.ai/docs/models-api/task-polling) (platform)

### [includeCost](https://runware.ai/docs/models/anthropic-claude-fable-5#request-includecost)

- **Type**: `boolean`

Include task cost in the response.

### [includeUsage](https://runware.ai/docs/models/anthropic-claude-fable-5#request-includeusage)

- **Type**: `boolean`

Include token usage statistics in the response.

### [numberResults](https://runware.ai/docs/models/anthropic-claude-fable-5#request-numberresults)

- **Type**: `integer`
- **Min**: `1`
- **Max**: `4`
- **Default**: `1`

Number of results to generate. Each result uses a different seed, producing variations of the same parameters.

**Inputs**

Input resources for the task (images, audio, etc). These must be nested inside the \`inputs\` object.

### [documents](https://runware.ai/docs/models/anthropic-claude-fable-5#request-inputs-documents)

- **Path**: `inputs.documents`
- **Type**: `array of strings`
- **Min items**: `1`
- **File size (each)**: `up to 4 MB`

Array of document inputs (UUID, URL, or Base64).

### [images](https://runware.ai/docs/models/anthropic-claude-fable-5#request-inputs-images)

- **Path**: `inputs.images`
- **Type**: `array of strings`
- **Min items**: `1`
- **File size (each)**: `up to 4 MB`

Array of image inputs (UUID, URL, Data URI, or Base64).

**Core Parameters**

Primary parameters that define the task output.

### [model](https://runware.ai/docs/models/anthropic-claude-fable-5#request-model)

- **Type**: `string`
- **Required**: true
- **Value**: `anthropic-claude-fable-5`

Send `auto` instead of an AIR identifier to have the model router choose the model for this task. The rest of the request is unchanged, every value you send is kept, and only the parameters you leave out are filled in. See [Model Router](https://runware.ai/docs/models-api/model-router).

**Learn more** (3 resources):

- [Text To Image: Model Selection](https://runware.ai/docs/learn/text-to-image#model-selection) (learn)
- [Image Inpainting: Model Specialized Inpainting Models](https://runware.ai/docs/learn/image-inpainting#model-specialized-inpainting-models) (learn)
- [Image Outpainting: Other Critical Parameters](https://runware.ai/docs/learn/image-outpainting#other-critical-parameters) (learn)

### [jsonSchema](https://runware.ai/docs/models/anthropic-claude-fable-5#request-jsonschema)

- **Type**: `object | string`

JSON Schema for structured output. Only honoured when `outputFormat` is JSON. Accepts the OpenAI envelope ({name, schema, strict}) or a bare JSON Schema; bare schemas are auto-wrapped with `name='response'` and `strict=true`.

### [messages](https://runware.ai/docs/models/anthropic-claude-fable-5#request-messages)

- **Path**: `messages.role`
- **Type**: `array of objects (2 properties)`
- **Required**: true
- **Min items**: `1`

Array of chat messages forming the conversation context. The final message must use the user role.

#### [role](https://runware.ai/docs/models/anthropic-claude-fable-5#request-messages-role)

- **Path**: `messages.role`
- **Type**: `string`
- **Required**: true

Role of the message author.

**Allowed values**: `user` `assistant`

#### [content](https://runware.ai/docs/models/anthropic-claude-fable-5#request-messages-content)

- **Path**: `messages.content`
- **Type**: `string`
- **Required**: true
- **Min**: `1`

Text content of the message.

**Settings**

Technical parameters to fine-tune the inference process. These must be nested inside the \`settings\` object.

### [systemPrompt](https://runware.ai/docs/models/anthropic-claude-fable-5#request-settings-systemprompt)

- **Path**: `settings.systemPrompt`
- **Type**: `string`
- **Min**: `1`
- **Max**: `200000`

System-level instruction that guides the model's behavior and output style across the entire generation.

### [cache](https://runware.ai/docs/models/anthropic-claude-fable-5#request-settings-cache)

- **Path**: `settings.cache`
- **Type**: `object (2 properties)`

Prompt caching configuration. Caches designated parts of the request to reduce cost and latency on repeated calls.

#### [scope](https://runware.ai/docs/models/anthropic-claude-fable-5#request-settings-cache-scope)

- **Path**: `settings.cache.scope`
- **Type**: `string`
- **Default**: `system+history`

Controls which parts of the request are cached.

**Allowed values**:

- `system` Cache the system prompt only.
- `system+history` Cache the system prompt and conversation history up to the last user message.

#### [ttl](https://runware.ai/docs/models/anthropic-claude-fable-5#request-settings-cache-ttl)

- **Path**: `settings.cache.ttl`
- **Type**: `string`
- **Default**: `5m`

Time-to-live for the cache.

**Allowed values**: `5m` `1h`

### [maxTokens](https://runware.ai/docs/models/anthropic-claude-fable-5#request-settings-maxtokens)

- **Path**: `settings.maxTokens`
- **Type**: `integer`
- **Min**: `1`
- **Max**: `128000`

Maximum number of tokens to generate in the response.

### [splitThinking](https://runware.ai/docs/models/anthropic-claude-fable-5#request-settings-splitthinking)

- **Path**: `settings.splitThinking`
- **Type**: `boolean`
- **Default**: `true`

When enabled, the model's internal reasoning is separated from the main response and returned in a dedicated `reasoningContent` field.

### [stopSequences](https://runware.ai/docs/models/anthropic-claude-fable-5#request-settings-stopsequences)

- **Path**: `settings.stopSequences`
- **Type**: `array of strings`
- **Min**: `1`
- **Max**: `50`
- **Max items**: `5`

Array of sequences that will cause the model to stop generating further tokens when encountered.

### [thinkingLevel](https://runware.ai/docs/models/anthropic-claude-fable-5#request-settings-thinkinglevel)

- **Path**: `settings.thinkingLevel`
- **Type**: `string`
- **Default**: `high`

Controls the depth of internal reasoning the model performs before generating a response.

**Allowed values**: `low` `medium` `high` `xhigh` `max` `adaptive`

### [toolChoice](https://runware.ai/docs/models/anthropic-claude-fable-5#request-toolchoice)

- **Path**: `toolChoice.type`
- **Type**: `object (2 properties)`

Controls how the model selects which tool to call. This only takes effect when `tools` are defined.

**Examples**:

**Let the model decide (default):**

```json
"toolChoice": {
  "type": "auto"
}
```

**Force a specific tool call:**

```json
"toolChoice": {
  "type": "tool",
  "name": "get_weather"
}
```

**Require any tool call:**

```json
"toolChoice": {
  "type": "any"
}
```

#### [type](https://runware.ai/docs/models/anthropic-claude-fable-5#request-toolchoice-type)

- **Path**: `toolChoice.type`
- **Type**: `string`
- **Required**: true

Strategy the model uses to decide when and which tools to call.

**Allowed values**:

- `auto` Tool calls decided by the model from the conversation context. This is the recommended default.
- `any` At least one tool call, with the model choosing which tool. Useful when you always need structured output.
- `tool` Call to the specific tool identified by `name`. Use this to force a particular function call.
- `none` No tool calls, even if tools are defined. Useful for forcing a text-only response.

#### [name](https://runware.ai/docs/models/anthropic-claude-fable-5#request-toolchoice-name)

- **Path**: `toolChoice.name`
- **Type**: `string`

Name of the specific tool the model must call. Required when type is `tool`.

### [tools](https://runware.ai/docs/models/anthropic-claude-fable-5#request-tools)

- **Path**: `tools.type`
- **Type**: `array of objects (4 properties)`
- **Min items**: `1`

An array of tool definitions that the model may call during generation. The model can invoke one or more tools based on the conversation context, outputting structured calls with arguments instead of (or alongside) free-text.

For `function` tools, each definition requires:

- **`type`**: `"function"`
- **`name`**: Unique identifier (alphanumeric, hyphens, underscores; max 64 chars).
- **`description`**: What the function does. The model uses this to decide when to call it.
- **`schema`**: JSON Schema object describing the expected input arguments.

> [!NOTE]
> The `search` tool is executed server-side by the provider. You don't need to handle the tool result yourself.

> [!NOTE]
> The `codeInterpreter` tool is executed server-side by the provider. You don't need to handle the tool result yourself.

**Examples**:

**Function tool, weather lookup:**

```json
"tools": [
  {
    "type": "function",
    "name": "get_weather",
    "description": "Get current weather for a city",
    "schema": {
      "type": "object",
      "properties": {
        "city": { "type": "string", "description": "City name" }
      },
      "required": ["city"]
    }
  }
],
"toolChoice": { "type": "auto" }
```

**Built-in web search:**

```json
"tools": [
  { "type": "search" }
]
```

**Built-in code interpreter:**

```json
"tools": [
  { "type": "codeInterpreter" }
]
```

**Multiple function tools:**

```json
"tools": [
  {
    "type": "function",
    "name": "search_products",
    "description": "Search the product catalog by query and filters.",
    "schema": {
      "type": "object",
      "properties": {
        "query": { "type": "string" },
        "category": { "type": "string" }
      },
      "required": ["query"]
    }
  },
  {
    "type": "function",
    "name": "add_to_cart",
    "description": "Add a product to the user's shopping cart.",
    "schema": {
      "type": "object",
      "properties": {
        "productId": { "type": "string" },
        "quantity": { "type": "integer", "minimum": 1 }
      },
      "required": ["productId"]
    }
  }
]
```

#### [type](https://runware.ai/docs/models/anthropic-claude-fable-5#request-tools-type)

- **Path**: `tools.type`
- **Type**: `string`
- **Required**: true

Kind of tool to make available to the model. User-defined functions require `name` and `schema`, while built-in tools (`search`, `codeInterpreter`) are executed server-side by the provider.

**Allowed values**: `function`

#### [name](https://runware.ai/docs/models/anthropic-claude-fable-5#request-tools-name)

- **Path**: `tools.name`
- **Type**: `string`
- **Max**: `64`

Unique function name. Required for function tools.

#### [description](https://runware.ai/docs/models/anthropic-claude-fable-5#request-tools-description)

- **Path**: `tools.description`
- **Type**: `string`

Explanation of what the function does, used by the model to decide when to call it.

#### [schema](https://runware.ai/docs/models/anthropic-claude-fable-5#request-tools-schema)

- **Path**: `tools.schema`
- **Type**: `object`

JSON Schema object describing the function's input parameters.

## Response Parameters

### [taskType](https://runware.ai/docs/models/anthropic-claude-fable-5#response-tasktype)

- **Type**: `string`
- **Required**: true
- **Value**: `textInference`

Identifier for the type of task this response belongs to.

### [taskUUID](https://runware.ai/docs/models/anthropic-claude-fable-5#response-taskuuid)

- **Type**: `string`
- **Required**: true
- **Format**: `UUID v4`

UUID v4 identifier echoed from the original request, used to match async responses to their tasks.

### [text](https://runware.ai/docs/models/anthropic-claude-fable-5#response-text)

- **Type**: `string`
- **Required**: true

Generated text content.

### [reasoningContent](https://runware.ai/docs/models/anthropic-claude-fable-5#response-reasoningcontent)

- **Type**: `string`

The model's internal reasoning content, separated from the main response. Only present when `settings.splitThinking` is enabled.

### [cost](https://runware.ai/docs/models/anthropic-claude-fable-5#response-cost)

- **Type**: `float`

Task cost in USD. Present when `includeCost` is set to `true` in the request.

### [finishReason](https://runware.ai/docs/models/anthropic-claude-fable-5#response-finishreason)

- **Type**: `string`
- **Required**: true

The reason why the model stopped generating tokens.

**Possible values**: `stop` `length` `content_filter` `unknown`

### [usage](https://runware.ai/docs/models/anthropic-claude-fable-5#response-usage)

- **Path**: `usage.promptTokens`
- **Type**: `object (4 properties)`
- **Required**: true

Token usage statistics for the request.

#### [promptTokens](https://runware.ai/docs/models/anthropic-claude-fable-5#response-usage-prompttokens)

- **Path**: `usage.promptTokens`
- **Type**: `integer`
- **Required**: true
- **Min**: `0`

Number of tokens in the input prompt.

#### [completionTokens](https://runware.ai/docs/models/anthropic-claude-fable-5#response-usage-completiontokens)

- **Path**: `usage.completionTokens`
- **Type**: `integer`
- **Required**: true
- **Min**: `0`

Number of tokens generated in the response.

#### [totalTokens](https://runware.ai/docs/models/anthropic-claude-fable-5#response-usage-totaltokens)

- **Path**: `usage.totalTokens`
- **Type**: `integer`
- **Required**: true
- **Min**: `0`

Total number of tokens used (prompt + completion).

#### [thinkingTokens](https://runware.ai/docs/models/anthropic-claude-fable-5#response-usage-thinkingtokens)

- **Path**: `usage.thinkingTokens`
- **Type**: `integer`
- **Min**: `0`

Number of tokens used for internal reasoning. Billed separately.

## Examples

### structured-output (structured-output)

```
```json
{
  "title": "Decision Brief: Deploy the Second Mobile Prenatal Clinic to Mesa Verde County",
  "summary": [
    "Applying the board's weighted priorities, Mesa Verde County is the recommended pilot site. It decisively outperforms North Fork on the two highest-weighted criteria, access inequity (40%) and maternal risk (25%): 31% of pregnant residents receive inadequate prenatal care versus 24% in North Fork, the preterm birth rate is higher (14.2% versus 11.8%), only two obstetric clinicians serve 1,180 annual births compared with five clinicians for 1,540 births, and the median trip to labor and delivery is 47 miles versus 33 miles. Mesa Verde also compares favorably on partner readiness (20%): the local federally qualified health center commits parking, specimen handling, referral coordination, and a 0.5-FTE bilingual navigator, a broader operational package than North Fork's weekly ultrasound access and 0.25-FTE nurse coordination.",
    "North Fork leads only on reach efficiency (15%), with a projected 15 completed visits per operating day versus 13, lower estimated weekly travel (290 versus 420 vehicle miles), and a lower annual operating cost ($452,000 versus $486,000). Both sites fit within the $500,000 budget ceiling, though Mesa Verde leaves less contingency headroom. Because the evidence is not close on need — Mesa Verde shows greater unmet need on every access and risk indicator — the board's tie-breaker instruction reinforces rather than decides the recommendation. All figures are planning estimates and should be validated before contracting."
  ],
  "recommendation": "Select Mesa Verde County for the 12-month, three-day-per-week mobile prenatal clinic pilot, contingent on executing a written agreement with the federally qualified health center and confirming that the $486,000 operating estimate holds within the $500,000 budget cap.",
  "key_points": [
    "Comparative need strongly favors Mesa Verde: 31% of pregnant residents receive inadequate prenatal care compared with 24% in North Fork, the preterm birth rate is 14.2% versus 11.8%, and only two obstetric clinicians serve the county versus five in North Fork, indicating substantially greater access inequity and maternal risk where the board's weights are heaviest.",
    "Expected reach modestly favors North Fork: at 15 completed visits per operating day and a 71% show rate, North Fork would deliver roughly 2,340 completed visits over 52 weeks of three-day service, compared with roughly 2,030 in Mesa Verde at 13 visits per day and a 76% show rate, though Mesa Verde's higher show rate suggests stronger per-appointment engagement.",
    "Partner readiness favors Mesa Verde because the federally qualified health center offers a fuller operational package — parking, specimen handling, referral coordination, and a 0.5-FTE bilingual navigator — whereas North Fork's county hospital offers one day per week of ultrasound access and 0.25 FTE of nurse coordination, a valuable but narrower commitment.",
    "Cost and logistics favor North Fork on paper: its $452,000 annual operating estimate is $34,000 lower than Mesa Verde's $486,000, and its 290 estimated weekly vehicle miles are well below Mesa Verde's 420; both options fit under the $500,000 ceiling, but Mesa Verde leaves only about $14,000 of contingency headroom.",
    "Equity considerations reinforce Mesa Verde: 62% of births are Medicaid-covered versus 54% in North Fork, the median 47-mile trip to labor and delivery is the longer of the two, and 28% of households primarily speak Spanish — a language need directly matched by the partner's bilingual navigator, whereas North Fork's 12% Hmong-speaking households have no identified language support in the current partner offer.",
    "Pre-launch validation is essential because all figures are planning estimates: the organization should confirm the visit-volume and show-rate projections against comparable operations, secure the FQHC's in-kind commitments in a signed memorandum, stress-test the $486,000 cost estimate and 420-mile travel assumption, and verify that the county's two obstetric clinicians can absorb referrals generated by the pilot."
  ],
  "principal_tradeoff": "Choosing Mesa Verde accepts lower reach efficiency — fewer projected completed visits, higher travel mileage, and roughly $34,000 more in annual operating cost with thin budget contingency — in exchange for serving the community with markedly greater unmet need, higher maternal risk, stronger partner support, and closer alignment with the board's equity-weighted priorities.",
  "assumptions_to_validate": [
    "The projected 13 completed visits per operating day and 76% show rate in Mesa Verde are achievable given local demand, road conditions, and seasonal factors.",
    "The federally qualified health center's commitments (parking, specimen handling, referral coordination, and the 0.5-FTE bilingual navigator) will be formalized, funded, and staffed for the full 12-month pilot.",
    "The $486,000 annual operating estimate is complete and stable enough to remain under the $500,000 cap given the limited contingency margin.",
    "The 420 vehicle miles per week travel estimate reflects realistic routing across the county's service area.",
    "Mesa Verde's two practicing obstetric clinicians and the regional labor-and-delivery facility can accept the referral volume the mobile clinic will generate.",
    "Language access needs beyond Spanish, and any Medicaid billing or licensure requirements for mobile services in the county, are confirmed before contracting."
  ]
}
```
```

Finish reason: `stop`

```json
{
  "taskType": "textInference",
  "taskUUID": "fab31eae-4609-4868-b06e-4e44110c6401",
  "outputFormat": "JSON",
  "model": "anthropic:claude@fable-5",
  "settings": {
    "splitThinking": false
  },
  "messages": [
    {
      "role": "user",
      "content": "Prepare an evidence-based decision brief for the board of a maternal-health nonprofit choosing where to deploy a second mobile prenatal clinic. The clinic will operate three days per week for a 12-month pilot. Compare the two finalist counties using the supplied facts, make a clear recommendation, explain the principal tradeoff, and identify assumptions that should be validated before contracting.\n\nDecision priorities: access inequity 40%, maternal risk 25%, partner readiness 20%, and reach efficiency 15%. If the evidence is close, favor the community with the greater unmet need. The annual operating budget cannot exceed $500,000.\n\nMesa Verde County: 1,180 births per year; 31% of pregnant residents receive inadequate prenatal care; 14.2% preterm birth rate; two practicing obstetric clinicians; median trip to labor and delivery is 47 miles; 62% of births are covered by Medicaid; 28% of households primarily speak Spanish. The local federally qualified health center will provide parking, specimen handling, referral coordination, and a 0.5-FTE bilingual navigator. The pilot is projected to deliver 13 completed visits per operating day at a 76% show rate. Estimated travel is 420 vehicle miles per week and annual operating cost is $486,000.\n\nNorth Fork County: 1,540 births per year; 24% receive inadequate prenatal care; 11.8% preterm birth rate; five practicing obstetric clinicians; median trip to labor and delivery is 33 miles; 54% of births are covered by Medicaid; 12% of households primarily speak Hmong. The county hospital will provide ultrasound access one day per week and 0.25 FTE of nurse coordination. The pilot is projected to deliver 15 completed visits per operating day at a 71% show rate. Estimated travel is 290 vehicle miles per week and annual operating cost is $452,000.\n\nTreat all figures as planning estimates rather than audited results. Do not invent external facts, additional statistics, or false precision. The title should state the recommended site. The summary should give the decision and rationale in two concise paragraphs suitable for a board packet. Provide six key points covering comparative need, expected reach, partner readiness, cost and logistics, equity considerations, and pre-launch validation. Make every key point a self-contained, informative statement rather than a fragment."
    }
  ],
  "jsonSchema": {
    "name": "response",
    "strict": true,
    "schema": {
      "type": "object",
      "properties": {
        "title": {
          "type": "string",
          "description": "Decision-brief title naming the recommended county."
        },
        "summary": {
          "type": "string",
          "description": "A concise two-paragraph executive summary of the recommendation, rationale, and principal tradeoff."
        },
        "keyPoints": {
          "type": "array",
          "description": "Six self-contained points covering the required decision factors.",
          "minItems": 6,
          "maxItems": 6,
          "items": {
            "type": "string"
          }
        }
      },
      "required": [
        "title",
        "summary",
        "keyPoints"
      ],
      "additionalProperties": false
    }
  }
}
```

---

### split-thinking (split-thinking)

```
Recommended award: Bid C
Five-year NPC: $728,000
Advantage over runner-up: $49,000
```

Finish reason: `stop`

```json
{
  "taskType": "textInference",
  "taskUUID": "1a7eb871-7d74-4697-bce7-02a23bfc0948",
  "model": "anthropic:claude@fable-5",
  "settings": {
    "splitThinking": true,
    "thinkingLevel": "high"
  },
  "messages": [
    {
      "role": "user",
      "content": "You are supporting a commercial real-estate procurement team selecting a rooftop solar contractor. Determine the bid with the lowest five-year net present cost (NPC).\n\nUse this formula:\nNPC = upfront price − signing rebate + PV of maintenance and replacements − PV of energy savings − PV of guaranteed credits.\n\nAll annual amounts are paid or received at year-end. Use these discount factors exactly:\nYear 1: 0.9259\nYear 2: 0.8573\nYear 3: 0.7938\nYear 4: 0.7350\nYear 5: 0.6806\n\nEnergy is worth $100 per MWh in every year.\n\nBid A:\n- Upfront price: $1,240,000\n- Signing rebate: $160,000\n- Maintenance: $14,000 annually in years 1–5\n- Generation: 900 MWh annually in years 1–5\n\nBid B:\n- Upfront price: $1,180,000\n- Signing rebate: $90,000\n- Maintenance: $8,000 annually in years 1–5\n- Inverter replacement: $70,000 in year 4\n- Generation: 950 MWh annually in years 1–5\n\nBid C:\n- Upfront price: $1,310,000\n- Signing rebate: $220,000\n- Maintenance: included in years 1–2, then $10,000 annually in years 3–5\n- Generation: 920 MWh annually in years 1–5\n- Guaranteed performance credit: $25,000 in year 5\n\nPerform the multi-step calculation carefully. In the final response, show no working and return exactly three lines:\nRecommended award: Bid [letter]\nFive-year NPC: $[amount rounded to nearest $1,000]\nAdvantage over runner-up: $[amount rounded to nearest $1,000]"
    }
  ]
}
```

---

### text-generation (text-generation)

```
# A Landmark Reopened. A Future Within Reach.

On a gray weekday afternoon in Harbor Point, rain streaks the limestone façade of the old Carnegie library and pools at the foot of its oak doors, locked now for six years. Behind the glass, the great barrel-vaulted reading room sits in shadow, its terrazzo floors dusty and still. School has just let out. Children pass under the eaves with backpacks and hoods pulled up, hurrying toward apartments where, in nearly one home out of three, no reliable internet waits for homework or a job search. The building watches them go, as it has watched this neighborhood for more than a century. But stand here a moment longer and you can see something else: light in the windows, doors propped open, a room full of voices. That future is close enough to touch.

## The Door That Closed

For 107 years, the Harbor Point branch was the one address in this former shipbuilding neighborhood that asked nothing of anyone who walked in. Built in 1912, it served generations of families through the rise and decline of the yards, through wars and recessions, through every change the waterfront could deliver. Then the roof failed—not once, but repeatedly—and in 2019 the city had no choice but to close the doors. The building has stood vacant ever since.

The cost of that closure is measurable. In the library's service area, home to 18,400 residents, local schools report that only 38 percent of third graders read at grade level. Thirty-one percent of neighborhood households lack a fixed broadband subscription, which means that applying for a job, attending a telehealth appointment, or finishing a school assignment often requires a bus ride out of the neighborhood—or simply doesn't happen. A library cannot solve these challenges alone. But it is hard to name another single institution that addresses so many of them at once, in one trusted place, at no cost to anyone who enters.

The Harbor Point Learning Trust believed the story of this building was not finished. The Trust acquired it from the city for one dollar, under a preservation covenant that ensures the landmark will be protected as it is renewed. Now the work of reopening it has begun in earnest.

## What Will Open Here

The restored library will hold 14,200 square feet of purpose. A children's reading room will give the neighborhood's youngest residents a bright, welcoming place to fall in love with books—the daily, unglamorous work behind every gain in early literacy. Six free tutoring rooms will let volunteers and students work side by side without distraction, session after session, year after year.

For the many households without home broadband, a 42-seat digital learning lab will offer fast, free access along with the guidance to use it well—for coursework, certifications, and the digital skills that today's employers assume. A small-business resource desk will support neighbors starting or growing enterprises of their own, from licensing questions to business plans. A quiet room, designed for privacy, will give residents a dignified place for telehealth appointments and job interviews—small moments on which health and livelihoods often turn. And a flexible community hall will host everything from civic meetings to family programs, restoring a gathering place this neighborhood has gone too long without.

The branch will operate six days a week in partnership with the city library system, and every core service will be free. This will not be a monument to what Harbor Point was. It will be working infrastructure for what Harbor Point is becoming.

## A Building Worth Carrying Forward

Some buildings can be replaced. This one should not be. The renewal design retains what generations of residents have loved: the limestone façade, the original oak doors, the terrazzo floors, and the barrel-vaulted ceiling that makes the main reading room feel like a civic sanctuary. Children who walk in beneath that ceiling in 2028 will stand in the same light their great-grandparents knew.

At the same time, the building will be made ready for its second century. An all-electric mechanical system will replace failing equipment with clean, efficient technology. High-performance interior storm windows will preserve the historic glass while dramatically improving comfort and efficiency. A new elevator and step-free entry will open every floor to every resident for the first time in the building's history. Outside, a rain garden will manage stormwater—a quiet answer to the very element that once forced these doors closed.

This is preservation with a purpose: honoring craftsmanship not as nostalgia, but as a foundation sturdy enough to carry the neighborhood's future.

## The Moment to Act

The total campaign goal is $8.6 million. Of that amount, $5.1 million is already committed—through state historic-tax-credit equity, a city capital grant, and two lead foundation gifts. That early confidence from public and philanthropic partners speaks to the strength of this project. What remains is $3.5 million, to be raised in this public phase, and it is the portion that only a community of donors can provide.

The timeline is set. Construction begins in March 2027. The library reopens in autumn 2028. Every gift made now moves directly toward that opening day—toward the first story hour in the children's reading room, the first tutoring session, the first neighbor who logs on in the learning lab to apply for a job.

Restore a landmark, reopen opportunity.

Those six words carry the whole of this campaign. The landmark is real: limestone and oak, terrazzo and vaulted plaster, a century of civic memory standing at the heart of Harbor Point. The opportunity is just as real: a third grader learning to read, a parent keeping a telehealth appointment with privacy and dignity, a founder getting a first business off the ground, a household finally connected to the wider world. Neither can be reopened without the other—and neither can be reopened without you.

We invite you to make a gift to the Harbor Point Library Renewal Campaign today, and to join the partners, foundations, and neighbors who have already brought this project more than halfway home. Your generosity will unlock the doors of a building that has waited long enough, for a neighborhood that has waited longer.

Give now, and help open these doors for good.
```

Finish reason: `stop`

```json
{
  "taskType": "textInference",
  "taskUUID": "b20e6a28-90d2-433d-a6d5-ce164b632c53",
  "model": "anthropic:claude@fable-5",
  "settings": {
    "splitThinking": false
  },
  "messages": [
    {
      "role": "user",
      "content": "Write the final donor-facing case statement for the Harbor Point Library Renewal Campaign. This must be a polished, self-contained piece ready for a campaign brochure—not an outline, memo, or commentary on the brief.\n\nProject facts:\n- Harbor Point is a former shipbuilding neighborhood in Baltimore with 18,400 residents.\n- Its 1912 Carnegie library closed in 2019 after repeated roof failures and has remained vacant.\n- The nonprofit Harbor Point Learning Trust acquired the building from the city for $1 under a preservation covenant.\n- The campaign goal is $8.6 million. Of that amount, $5.1 million is already committed through state historic-tax-credit equity, a city capital grant, and two lead foundation gifts. The public phase must raise the remaining $3.5 million.\n- The restored 14,200-square-foot building will include a children’s reading room, a 42-seat digital learning lab, six free tutoring rooms, a flexible community hall, a small-business resource desk, and a quiet room for telehealth appointments and job interviews.\n- The branch will operate six days per week in partnership with the city library system. All core services will be free.\n- Local schools report that 38% of third graders in the service area read at grade level. Thirty-one percent of neighborhood households lack a fixed broadband subscription.\n- Preservation work will retain the limestone façade, original oak doors, terrazzo floors, and the main reading room’s barrel-vaulted ceiling.\n- The design includes an all-electric mechanical system, a new elevator, step-free entry, high-performance interior storm windows, and a rain garden.\n- Construction is scheduled to begin in March 2027 and the library is expected to reopen in autumn 2028.\n- Campaign message: restore a landmark, reopen opportunity.\n\nRequirements:\n- Write 900–1,100 words.\n- Use the title “A Landmark Reopened. A Future Within Reach.”\n- Begin with a vivid 100–140-word opening that places the reader outside the shuttered library on a rainy weekday afternoon, then turns toward possibility. Do not invent a named resident or quote.\n- Organize the remainder under exactly these four subheads: “The Door That Closed,” “What Will Open Here,” “A Building Worth Carrying Forward,” and “The Moment to Act.”\n- Make the case emotionally compelling while remaining credible, specific, and restrained. Use warm, lucid language suitable for individual donors, foundations, and civic partners.\n- Include the campaign figures clearly and explain that $5.1 million is committed and $3.5 million remains to be raised.\n- Connect literacy, broadband access, workforce needs, and preservation without presenting the library as a cure-all.\n- Mention every planned program space at least once, but do not turn the prose into a feature list.\n- Include one short standalone sentence using the campaign message verbatim: “Restore a landmark, reopen opportunity.”\n- End with a direct invitation to give, followed by a final sentence of no more than 12 words.\n- Do not add donor tiers, naming opportunities, testimonials, fabricated quotations, footnotes, citations, contact details, or facts not supplied here.\n- Output only the finished case statement."
    }
  ]
}
```

---

### reference-to-text (reference-to-text)

```
# Preliminary Freight Damage Assessment — Crated Equipment, Receiving Dock

**Report type:** Preliminary visual survey (single photograph)
**Basis:** One receiving-dock photograph provided by the equipment owner. No shipment documents, labels, or physical access available to the surveyor at time of writing.
**Shipment identifiers:** Not established. Labels visible in the photograph are not legible at the provided resolution; no identifiers are asserted in this report.

---

## 1. Executive Summary

The photograph shows a crated appliance-style unit — consistent in form with a piece of laboratory or light industrial equipment (drum/bowl visible at top, control panel on the upper front face) — on a wooden pallet inside what appears to be a warehouse or receiving-dock environment.

The shipment presents **significant, readily visible transit damage**:

- The plywood crate is **broken open**, with a large torn/splintered plywood panel resting across the top of the unit and the front and left enclosure panels missing or displaced.
- The equipment's front/left corner panel shows a **substantial impact deformation**: torn sheet metal, crushed panel edges, and exposed foam or residue around the puncture area.
- The pallet deck is **fractured**, with multiple broken slats and loose wood debris scattered on the floor in the foreground.

**Apparent severity:** Major. The protective packaging failed and the equipment enclosure itself has been breached at at least one location. Internal damage is plausible but **cannot be confirmed** from this photograph.

**Immediate priority:** Preserve the shipment exactly as found, document thoroughly, note the exception on the delivery receipt (if not yet signed clean), and notify the carrier in writing without delay. **Do not power, operate, or fully unpack the unit** before a formal inspection.

---

## 2. Visible Evidence Table

| # | Location in Image | Observed Condition | Significance | Confidence |
|---|---|---|---|---|
| 1 | Upper center/right, resting atop unit | Large plywood panel with a jagged, splintered torn edge, lying diagonally across the top of the equipment | Crate lid or side panel failed under force; indicates the crate absorbed a significant load or impact | High |
| 2 | Left side of frame | Angled plywood panel leaning against the unit rather than fastened vertically | Crate left wall displaced or detached; enclosure integrity lost | High |
| 3 | Right side of frame | Intact plywood crate wall with two horizontal gray/metal bands still in place | Portion of crate remained secured; suggests force was localized or directional rather than total collapse | High |
| 4 | Center, front/right corner of equipment (mid-height) | Torn, crumpled sheet metal; panel edge split open; deformation of the cabinet corner | Direct impact or crush damage to the equipment enclosure itself, not just packaging | High |
| 5 | Center front panel, around and below the breach | Foam-like material and irregular staining/residue on the painted surface; drip-pattern marks running downward | Possible exposed internal insulation, escaped packing foam, or leaked substance; nature of residue not determinable from photo | Medium |
| 6 | Upper left of unit | Clear plastic sheeting/wrap bunched and detached, no longer covering the control panel area | Protective wrap failed or was disturbed; control panel and top surfaces exposed | High |
| 7 | Upper center of unit | Stainless-style circular drum/bowl rim visible at top; control panel with display visible beneath wrap | Identifies general equipment type and shows sensitive surfaces exposed; no obvious cracking visible on these surfaces at this resolution | Medium |
| 8 | Lower center, pallet deck | Multiple broken/fractured deck boards; splintered slats beneath and beside the equipment feet | Pallet structurally compromised; unit may be inadequately supported; handling hazard | High |
| 9 | Foreground floor, lower left and center | Scattered wood fragments and splinters on the concrete floor | Breakage occurred at or before this location; debris field is evidence and a slip/puncture hazard | High |
| 10 | Center-right of equipment and lower-right pallet foot | White labels affixed to equipment and to pallet | Potentially contain serial numbers, shipment or handling information — **text is not legible in this photograph** | High (presence); N/A (content) |
| 11 | Center-right of unit | Small blue logo/badge on front panel | Manufacturer branding present but not legible at this resolution | Medium |

---

## 3. Packaging and Load-Security Assessment

**Crate:** The plywood crate has suffered partial structural failure. The right-hand wall remains upright and banded, but the top/front panel is torn with a jagged fracture edge and is lying loose across the equipment, and the left panel is leaning at an angle rather than fastened. The tear pattern on the loose panel (splintered, irregular edge rather than a clean saw cut) is consistent with mechanical failure under force rather than deliberate opening, though this is an interpretation, not a certainty.

**Bands/strapping:** Two horizontal bands are visible and appear intact on the surviving right wall. No banding is visible restraining the equipment on the open (left/front) side, meaning the unit is currently **not fully secured** within the remains of the crate.

**Wrapping:** The clear plastic protective wrap is torn/displaced and bunched at the upper left, leaving the control panel and upper surfaces exposed to dust, moisture, and contact.

**Pallet:** Multiple deck boards are fractured directly beneath and adjacent to the equipment feet, with debris on the floor. The pallet's load-bearing capacity is compromised; forklift or pallet-jack movement in this state risks further damage or a drop.

**Displacement:** The equipment appears to still be sitting roughly within the pallet footprint; no evidence of the unit having fully fallen off the pallet is visible, though a partial shift or tip-and-recover event cannot be ruled out from a single angle.

---

## 4. Equipment Condition Assessment (Visible Surfaces Only)

- **Front/right corner panel (mid-height):** The most significant visible equipment damage. Sheet metal is torn open and crumpled at a panel seam/corner, with the surrounding panel deformed. This is a breach of the enclosure, not superficial scuffing.
- **Front panel surface:** Staining and residue with downward drip patterns below the breach, plus a foam-like material at the tear. Whether this is packaging foam, internal insulation, or a leaked fluid **cannot be determined from the photograph**.
- **Control panel (upper front):** Visible through/behind displaced wrap. No cracks or shattered display elements are clearly identifiable at this resolution, but the panel cannot be cleared of damage from this image.
- **Top drum/bowl assembly:** The visible rim appears intact from this angle; the interior and far side are not visible.
- **Feet/base:** The unit's leveling feet are visible resting on broken pallet boards; alignment or bending of the feet cannot be assessed from this angle.
- **Rear and right side of unit, internal components:** **Not visible; condition unknown.** Given the enclosure breach and packaging failure, concealed internal damage (mechanical, electrical, alignment) is plausible and must be assumed possible until inspected.

---

## 5. Possible Damage Mechanisms (Hypotheses — Not Conclusions)

The following are ranked plausible mechanisms based solely on visible evidence. None is established as fact.

**1. Forklift or handling-equipment strike (most consistent with evidence).**
A localized, torn puncture in the equipment's corner panel, a fractured crate wall with a jagged break, and broken pallet boards concentrated in one area are all consistent with a concentrated horizontal or angled impact, such as fork tine or handling-equipment contact. The surviving intact, banded right wall suggests the force was directional rather than uniform.

**2. Drop or tip-over during transit or transfer.**
The broken pallet deck boards and the crate lid torn free could result from the crated unit being dropped or tipping and being righted. The debris field on the floor is consistent with breakage occurring at or near this location, though debris could also have fallen out during placement. Less strongly supported than #1 because the unit appears roughly upright and centered on the pallet.

**3. Crush/compression from stacked or shifting cargo.**
The failed top panel resting across the unit could indicate a top-down load (e.g., cargo stacked on or shifting onto the crate in transit). However, the pronounced side-panel puncture on the equipment is less typical of pure top-down compression, so this mechanism alone does not explain all the evidence.

A formal cause determination requires the additional evidence listed in Section 7 and, likely, a joint carrier inspection.

---

## 6. Immediate Preservation Actions

Receiving staff should, **without moving, opening, energizing, or further disturbing the shipment**:

1. **Do not sign a clean delivery receipt.** If the receipt has not yet been signed, note "Received damaged — crate broken open, equipment panel punctured, pallet broken, subject to inspection" and photograph the annotated receipt. If already signed clean, document the discovery time immediately in writing.
2. **Leave everything in place**, including the loose plywood panels, displaced wrap, and floor debris. Do not sweep up the wood fragments until photographed and, ideally, retained.
3. **Cordon off the area** with cones or tape. The broken pallet, splintered wood, and torn sheet metal edges are **laceration, puncture, and trip/slip hazards** — flag these to staff.
4. **Do not connect, energize, or test the equipment** under any circumstances before inspection, particularly given the enclosure breach and unidentified residue.
5. **Do not touch or attempt to identify the residue/foam** on the front panel until its nature is established; treat it as unknown material.
6. **Protect from further exposure** only by non-invasive means (e.g., loosely draping a tarp over the top if the dock is exposed to weather), without repositioning packaging components.
7. **Photograph immediately** per Section 7 while conditions are unchanged.
8. **Record chain of custody:** who received it, from which carrier/driver, arrival time, and who has approached it since.
9. **Retain all packaging** — crate panels, bands, wrap, pallet, and debris — until the claim is resolved or the carrier/insurer waives salvage.

---

## 7. Additional Evidence Required

**Photographs (high resolution, with a scale reference such as a tape measure):**
- All four sides of the crate/unit, plus top-down if safely possible.
- Close-ups of: the equipment panel breach and tear edges; the residue/staining; the fractured plywood edges; each broken pallet board; the intact banding; the floor debris field in situ.
- Legible close-ups of **every label** on the equipment, crate, and pallet (currently unreadable), including serial/model plates.
- Any shock, tilt, or temperature indicators on the crate, if fitted (none identifiable in this photograph).
- The delivery vehicle and adjacent cargo, if still on site.

**Documents:**
- Bill of lading / delivery receipt with the damage exception noted.
- Packing list, commercial invoice, and the shipper's packing/photographs at origin (to establish pre-shipment condition).
- Carrier's driver name, trailer/PRO number, and route information.

**Measurements:**
- Height of the impact/puncture above floor and pallet deck (relevant to fork-tine hypothesis).
- Dimensions of the panel breach and of the crate fracture.

**Inspection steps:**
- Formal joint inspection with the carrier's representative before any unpacking.
- Manufacturer or qualified technician assessment of internal mechanical/electrical condition, including drum/rotor alignment, **before any power-up**.
- Identification of the residue material.

---

## 8. Carrier Notification Narrative (Draft)

> We are providing formal notice of damage to a crated equipment shipment received at our dock. Upon arrival, the plywood crate was found broken open: the top/front panel was torn free with splintered edges and was resting loose on the unit, and a side panel was detached and leaning against the equipment. The equipment itself has sustained visible damage, including a torn and crushed sheet-metal panel at one front corner, exposed foam/residue and staining on the front surface, and displaced protective wrapping. The wooden pallet is fractured, with multiple broken deck boards and wood debris on the floor. The damage was documented photographically at the time of discovery, and all packaging materials have been preserved in place. The unit has not been moved, opened, or powered. We consider concealed internal damage possible and are requesting a joint inspection at the earliest opportunity. All rights under the applicable bill of lading and governing regulations are expressly reserved.

*(Approx. 160 words. Insert shipment identifiers, dates, and contact details from your records before sending.)*

---

## 9. Limitations and Recommended Escalation

**Limitations:**
- This assessment is based on a **single photograph from one angle**. The rear, right interior face, underside, and all internal components are not visible.
- Label text is **not legible**; no shipment identifiers, model numbers, or handling-indicator statuses could be verified.
- The nature of the residue/foam on the front panel is **undetermined**.
- No cause of damage, timeline, or party responsibility is established herein; Section 5 mechanisms are hypotheses only.
- No cost, dimension, or internal-damage estimates are offered, as none can be supported by this evidence.

**Recommended escalation:**
1. Immediate written notice to the carrier (Section 8) within the timeframes required by the bill of lading and applicable regulations.
2. Prompt notice to the cargo insurer with this report and the full photo set.
3. Request a **joint survey** with carrier and insurer representatives before unpacking.
4. Engage the equipment manufacturer or an authorized technician for internal inspection and a repair-versus-replace determination — **prior to any energization**.
5. If the residue is suspected to be a hazardous or unknown fluid, involve site EHS personnel before further handling.

---

*Prepared as a preliminary visual survey only. Findings are subject to revision upon physical inspection and receipt of shipment documentation.*
```

Finish reason: `stop`

```json
{
  "taskType": "textInference",
  "taskUUID": "83b89a8b-6f95-419e-8235-9d0bdf218c27",
  "model": "anthropic:claude@fable-5",
  "settings": {
    "splitThinking": false
  },
  "inputs": {
    "images": [
      "https://assets.runware.ai/assets/inputs/23984bd8-69c5-4848-bda9-6decee5d7f27.jpg"
    ]
  },
  "messages": [
    {
      "role": "user",
      "content": "Act as a senior freight-claims surveyor. Analyze the attached receiving-dock photograph and produce a detailed preliminary freight damage assessment suitable for an equipment owner, insurer, and carrier.\n\nGround every finding in visible evidence. Do not invent shipment identifiers, label text, internal damage, dates, costs, dimensions, or events that the photograph cannot establish. Clearly separate direct observations from plausible interpretations, and use uncertainty language where appropriate.\n\nWrite the report in professional Markdown with these sections:\n\n1. Executive summary: shipment condition, apparent severity, and immediate priority.\n2. Visible evidence table with columns for item number, location in image, observed condition, significance, and confidence. Use precise spatial references such as upper right, foreground, or exposed side panel.\n3. Packaging and load-security assessment covering the crate, pallet, bands, wrapping, and any visible displacement.\n4. Equipment condition assessment limited strictly to visible surfaces.\n5. Possible damage mechanisms: rank up to three plausible mechanisms, explain the visible evidence supporting each, and explicitly state that these are hypotheses rather than conclusions.\n6. Immediate preservation actions that receiving staff can take without moving, energizing, opening, or altering the shipment unnecessarily.\n7. Additional evidence required, including recommended photographs, documents, measurements, and inspection steps needed to support a formal claim.\n8. Concise carrier-notification narrative of 120–180 words that can be pasted into an initial notice of damage.\n9. Limitations and recommended escalation.\n\nFlag any safety hazards visible in the scene. If text on labels is unreadable, say so rather than guessing. Do not recommend operating or powering the equipment before inspection."
    }
  ]
}
```