new·The score now tells you which way it movedA brain's exam only ever grows: its own material writes questions, and so does every question a real caller asked and did not get answered. The score is a percentage over that growing set, so a brain that learned more could post a smaller number — and this week three did. One of them answered two MORE questions than the week before and showed eighteen points less. Printed as a single percentage, that reads as decline to a reader and as punishment to anyone who contributes material.all news →
mozg.beta
Sign in

AI SDK · Cookbook · all subjects

model capabilities

9 notes, read out of this brain and free to use. Each one was extracted from a source and is re-checked against its exam.

Accessing reasoning tokens from OpenAI response metadata

When using OpenAI reasoning models, you can access reasoning tokens from the response via `result.finalStep.providerMetadata?.openai?.reasoningTokens`.

OpenAI reasoningEffort values and behavior

The `reasoningEffort` option for OpenAI reasoning models controls how much internal reasoning the model performs. Values are: 'none' (no reasoning, GPT-5.1 models only), 'minimal' (bare-minimum reasoning), 'low' (fast, concise reasoning), 'medium' (balanced, default), 'high' (thorough reasoning), 'xhigh' (maximum reasoning, GPT-5.1-Codex-Max only). Lower values are faster and cheaper; higher values produce more thorough answers.

OpenAI reasoningSummary option for accessing model's thought process

The `reasoningSummary` option surfaces the OpenAI model's thought process when working with reasoning models. When `reasoningEffort` is set to a value other than 'none', the OpenAI Responses provider defaults `reasoningSummary` to 'detailed'. Set `reasoningSummary: null` to omit reasoning summaries. Values are 'auto' (condensed summary of reasoning) or 'detailed' (comprehensive reasoning output).

OpenAI textVerbosity option controls response length independently

The `textVerbosity` option controls the length and detail of OpenAI model's text response independently of reasoning. Values are 'low' (terse, minimal responses), 'medium' (balanced detail, default), 'high' (verbose, comprehensive responses).

Anthropic thinking (extended reasoning) with budgetTokens

Anthropic's thinking feature gives Claude models a dedicated thinking phase before responding. Enable it by providing a `thinking` object with `type: 'enabled'` and `budgetTokens` set to an upper limit on reasoning tokens. The `budgetTokens` value sets how many tokens the model can use for internal reasoning; higher budgets allow deeper reasoning but increase latency and cost. Thinking is supported on claude-opus-4-20250514, claude-sonnet-4-20250514, and claude-sonnet-4-5-20250929 models.

Anthropic effort option for controlling reasoning depth

The `effort` option for Anthropic provides a simpler way to control reasoning depth without specifying a token budget. It affects thinking, text responses, and function calls. Values are 'low' (minimal reasoning, fastest responses), 'medium' (balanced reasoning), 'high' (thorough reasoning, default).

Anthropic speed option for claude-opus-4-6 fast mode

For claude-opus-4-6, the `speed` option enables approximately 2.5x faster output token speeds. Values are 'fast' or 'standard'.

Accessing reasoning text from Anthropic generateText response

When using Anthropic generateText with reasoning enabled, the result includes `reasoning` and `reasoningText` properties containing the model's thought process.

Image models and supported sizes/aspect ratios

Image model support varies by provider: xAI Grok grok-imagine-image supports aspect ratios (1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 2:1, 1:2, 19.5:9, 9:19.5, 20:9, 9:20, auto). OpenAI dall-e-3 supports sizes (1024x1024, 1792x1024, 1024x1792). OpenAI dall-e-2 supports sizes (256x256, 512x512, 1024x1024). Amazon Bedrock amazon.nova-canvas-v1:0 supports sizes 320-4096 (multiples of 16) and aspect ratios 1:4 to 4:1 with max 4.2M pixels. Fal models (flux/dev, flux-lora, fast-sdxl, flux-pro/v1.1-ultra, ideogram/v2, recraft-v3, stable-diffusion-3.5-large, hyper-sdxl) all support aspect ratios (1:1, 3:4, 4:3, 9:16, 16:9, 9:21, 21:9). DeepInfra stabilityai/sd3.5 supports aspect ratios (1:1, 16:9, 1:9, 3:2, 2:3, 4:5, 5:4, 9:16, 9:21). DeepInfra FLUX models support sizes 256-1440 (multiples of 32). Google Imagen 4.0 models support aspect ratios (1:1, 3:4, 4:3, 9:16, 16:9). Replicate recraft-ai/recraft-v3 supports multiple specific sizes. Together.ai models support sizes (512x512, 768x768, 1024x1024). Black Forest Labs Flux models support aspect ratios from 3:7 (portrait) to 7:3 (landscape).

Give your agent this brain