AI Model Compare

Compare API prices, capabilities, and starter prompts for GPT, Claude, Gemini, Grok, and Muse. Pick up to five models to see where each fits.

Side-by-side comparison

5 of 5 models selected.

OpenAI

GPT-6.2 Sol

gpt-6.2-sol

API price / 1M tokens

Input
$2.00
Output
$10.00

$4.00 in and $15.00 out above 272K input tokens.

Capabilities

Token limits
1.05M context · 128K max output
Inputs → output
Text and images → text

Function calling, structured output, web search, and computer use via the Responses API.

Best for

Everyday coding, document work, and tool-using workflows.

Why it fits. It offers the same published context window as Astra at a quarter of the base token rates. That makes it practical for work you repeat often.

How to prompt it

State who the output is for and what form it should take. Include a representative example or an acceptance test.

Turn these release notes into a 150-word update for customers. Lead with user-visible changes and quote the source line for any breaking change.

Anthropic

Claude Sonnet 5.5

claude-sonnet-5-5

API price / 1M tokens

Input
$2.00
Output
$10.00

Standard Claude API rate.

Capabilities

Token limits
1M context · 128K max output
Inputs → output
Text and images → text

Adaptive thinking, tool use, and structured output.

Best for

Code review, collaborative writing, and document synthesis.

Why it fits. It combines a large context window with Anthropic’s fast latency tier, which helps when you need several passes over a codebase or draft.

How to prompt it

Separate instructions from source material. Give a target tone or review rubric and say which details must be preserved.

Review the attached change against the requirements. List each issue with file and line, user impact, and a small fix. Prioritize correctness and accessibility.

Google

Gemini 4 Hydrogen

gemini-4-hydrogen

API price / 1M tokens

Input
$1.50
Output
$7.50

Standard paid rates. Google also lists a free tier.

Capabilities

Token limits
1.05M shared context · 65.5K synchronous output · up to 1.05M with long decode
Inputs → output
Text, images, video, audio, and documents → text

Multimodal input, tool calling, and long decode.

Best for

Everyday coding, media analysis, and routine tool workflows.

Why it fits. It retains the shared Gemini 4 context and long-decode option at a lower listed paid rate than Carbon or Argon.

How to prompt it

Give a single concrete deliverable, relevant inputs, and an example.

From this demo recording, draft the five setup steps and give the timestamp for each.

xAI

Grok 4.7

grok-4.7

API price / 1M tokens

Input
$2.00
Output
$6.00

$4.00 in and $12.00 out at 200K context tokens or more.

Capabilities

Token limits
500K context
Inputs → output
Text and images → text

Configurable reasoning, structured output, and optional web and X Search tools.

Best for

Research on fast-moving topics and multi-step technical work.

Why it fits. With X Search enabled, it can surface original posts alongside web sources. You can combine recent social activity with published documentation and keep the evidence distinct.

How to prompt it

Give a time window and source rules. Ask for direct links and dates and have it distinguish official statements from reactions.

Using Web and X Search, summarize the last 48 hours of official release updates. Link each announcement and label unverified claims separately.

Meta

Muse Spark 1.3

muse-spark-1.3

API price / 1M tokens

Input
$1.25
Output
$4.25

Standard mode. The separate Contributor rate has different data terms.

Capabilities

Token limits
1.05M context
Inputs → output
Text, images, video, audio, PDF → text

Configurable reasoning, function calling, structured output, and web grounding.

Best for

Tool-connected assistants, long-context reasoning, and software work at scale.

Why it fits. It can combine multiple input formats and call tools at a lower base rate than Muse Wonder. That can reduce cost when a workflow makes several model calls per task.

How to prompt it

State the goal, allowed tools, and when to stop. Tell it what source to use when two results disagree.

Use the inventory tool to reconcile this order against the attached packing slip. Return mismatches with source references and mark any unavailable records for review.

How to read the comparison

Prices are in USD per million tokens for standard, uncached API requests. The input price covers what you send and the output price covers what the model generates, including billable reasoning tokens where applicable. Caching, batches, subscriptions, and tool calls can change the total.

Context is the amount of information a model can handle in one request. Where a provider gives a separate input limit, the card labels it as max input. Tools such as live search need to be enabled, and availability can vary by API or plan.

Gemini 3 and 4 accept PDF, Microsoft Office, and OpenDocument files. The Gemini API can also restrict processing to selected pages, spreadsheet ranges, or time intervals, so you pay input-token charges for the processed slice. See Gemini file processing.

In the Gemini Interactions API, Gemini 3.6 Flash, 3.8 Flash, and the Gemini 4 models can use the tool retrieval preview. It searches a tool library and loads relevant function schemas on demand. See Gemini tool retrieval.

OpenAI is winding down the Flex processing beta. Existing supported models can still use Flex while OpenAI finalizes the retirement date. GPT‑6.2 Astra and Sol support discounted background completion windows for eligible requests. See OpenAI processing options.

Grok Build 0.1 retired from the xAI API on November 19, 2026. The Lower cost preset now uses Grok 4.3. See the xAI retirement notice.

Use-case notes and starter prompts are editorial examples based on the capabilities each provider documents. The catalog covers general-purpose text-output models. Prices and model specifications checked .