Compare

Two or three models, side by side.

Rates, worked costs, limits and capabilities from the same data as every model page. Among the models you pick, the lowest price and the largest window are marked.

The comparison is in the address bar, so you can share it. Looking for something else? Browse all 34 models.

Tokens per day, by model

Last 7 days, UTC days

Real tokens served through this gateway, updated every 5 minutes. Ranked by usage, not quality.

Full ranking insights

Model 1Model 2Model 3 (optional)

Showing a default comparison. Picking other models needs JavaScript; every model's rates are also on the models page.

Model	Claude Opus 5.5Anthropic · FlagshipDiscount −30%	GPT-6.1 SolOpenAI · FlagshipDiscount −65%New	Kimi K3Moonshot AI · FlagshipDiscount −88%New	Claude Sonnet 5Anthropic · Balanced	Claude Haiku 4.5Anthropic · Fast	GPT-6 AstraOpenAI · FlagshipDiscount −65%New	GPT-5.6 SolOpenAI · FlagshipUnavailable	GPT-5.6 TerraOpenAI · BalancedUnavailable	Grok 4.6xAI · BalancedUnavailable	Claude Sonnet 5.5Anthropic · BalancedDiscount −30%New	Grok 4.7xAI · FlagshipDiscount −30%New	DeepSeek V4 FlashDeepSeek · FastDiscount −86%New	DeepSeek V4 ProDeepSeek · BalancedDiscount −50%New	Qwen3.8 MaxQwen · FlagshipDiscount −71%New	GLM 5.3Z.ai · BalancedDiscount −82%New	MiniMax M3MiniMax · BalancedDiscount −46%New	GLM 5.3 FlashZ.ai · FastDiscount −58%New	Qwen3.8 27BQwen · FastFreeNew	Qwen3.8 27B GroqQwen · FastFreeNew	Claude Fable 5.1 (free lane)Anthropic · FlagshipNew	Claude Opus 5.5 (free lane)Anthropic · FlagshipNew	Claude Sonnet 5.5 (free lane)Anthropic · BalancedNew	DeepSeek V4 Flash (free lane)DeepSeek · FastNew	DeepSeek V4 Pro (free lane)DeepSeek · BalancedNew	StepAudio 3 Chat PreviewStepFun · FastFreeNew	Claude Fable 5.1Anthropic · FlagshipDiscount −30%New	Gemini 4 ArgonGoogle · FlagshipDiscount −73%New	Muse WonderMeta · FlagshipUnavailableDiscount −30%New	MiMo V2.6 ProXiaomi · FlagshipDiscount −30%New	Step 5 PreviewStepFun · FlagshipDiscount −30%New	Claude Haiku 5.5Anthropic · FastDiscount −30%New	Muse Spark OpenMeta · BalancedDiscount −30%New	Mistral Large 4Mistral AI · BalancedDiscount −65%New	GPT-6 LunaOpenAI · FastDiscount −65%New
Price per 1M tokens
Input	~~Regular rate $4.00~~now $2.80	~~Regular rate $2.00~~now $0.70	~~Regular rate $3.00~~now $0.353Lowest	$2.00	$1.00	~~Regular rate $10.00~~now $3.50	$4.00	$2.00	$2.00	~~Regular rate $2.00~~now $1.40	~~Regular rate $2.00~~now $1.40	~~Regular rate $0.15~~now $0.0392	~~Regular rate $0.66~~now $0.49	~~Regular rate $2.00~~now $0.577	~~Regular rate $1.40~~now $0.396	~~Regular rate $0.30~~now $0.161	~~Regular rate $0.15~~now $0.07	$0.00	$0.00	$7.00	$2.80	$1.40	$0.0392	$0.49	$0.00	~~Regular rate $10.00~~now $7.00	~~Regular rate $2.00~~now $0.525	~~Regular rate $4.00~~now $2.80	~~Regular rate $0.435~~now $0.3045	~~Regular rate $1.00~~now $0.70	~~Regular rate $0.10~~now $0.07	~~Regular rate $1.25~~now $0.875	~~Regular rate $1.36~~now $0.476	~~Regular rate $0.10~~now $0.035
Output	~~Regular rate $20.00~~now $14.00	~~Regular rate $10.00~~now $3.50Lowest	~~Regular rate $15.00~~now $5.67	$10.00	$5.00	~~Regular rate $50.00~~now $17.50	$20.00	$12.00	$6.00	~~Regular rate $10.00~~now $7.00	~~Regular rate $6.00~~now $4.20	~~Regular rate $0.60~~now $0.0784	~~Regular rate $1.98~~now $0.98	~~Regular rate $6.00~~now $1.73	~~Regular rate $4.40~~now $0.77	~~Regular rate $1.20~~now $0.672	~~Regular rate $0.50~~now $0.21	$0.00	$0.00	$35.00	$14.00	$7.00	$0.0784	$0.98	$0.00	~~Regular rate $50.00~~now $35.00	~~Regular rate $10.00~~now $2.625	~~Regular rate $20.00~~now $14.00	~~Regular rate $0.87~~now $0.609	~~Regular rate $2.70~~now $1.89	~~Regular rate $0.50~~now $0.35	~~Regular rate $4.25~~now $2.975	~~Regular rate $4.18~~now $1.428	~~Regular rate $0.50~~now $0.175
Cached input	$0.20	$0.10Lowest	$0.30	$0.20	$0.10	$1.00	$0.40	$0.20	$0.50	$0.10	$0.50	$0.003	$0.022	$0.25	$0.26	$0.06	$0.03	Not offered	Not offered	Not offered	Not offered	Not offered	Not offered	Not offered	Not offered	$0.25	$0.10	$0.20	$0.0036	$0.10	$0.01	$0.15	$0.14	$0.01
Output vs input	5×	5×	16.1×	5×	5×	5×	5×	6×	3×	5×	3×	2×	2×	3×	1.9×	4.2×	3×	—	—	5×	5×	5×	2×	2×	—	5×	5×	5×	2×	2.7×	5×	3.4×	3×	5×
Example cost
1,000 chat replies (1.5k in, 400 out)	$9.80	$2.45Lowest	$2.798	$7.00	$3.50	$12.25	$14.00	$7.80	$5.40	$4.90	$3.78	$0.09	$1.128	$1.558	$0.902	$0.51	$0.19	$0	$0	$24.50	$9.80	$4.90	$0.09	$1.128	$0	$24.50	$1.838	$9.80	$0.70	$1.806	$0.246	$2.502	$1.286	$0.122
1M input + 200k output	$5.60	$1.40Lowest	$1.487	$4.00	$2.00	$7.00	$8.00	$4.40	$3.20	$2.80	$2.24	$0.05488	$0.686	$0.923	$0.55	$0.2954	$0.112	$0	$0	$14.00	$5.60	$2.80	$0.05488	$0.686	$0	$14.00	$1.05	$5.60	$0.4263	$1.078	$0.14	$1.47	$0.7616	$0.07
Limits
Context window	1M	1.05MLargest	1.05M	1M	200k	1.05M	1.05M	1.05M	500k	1M	500k	1M	1M	1M	1.02M	1M	Not published	204k	131.1k	1M	1M	1M	1M	1M	Not published	1M	1M	1.05M	1.05M	1M	1M	1.05M	524.3k	1.05M
Max output	128k	128k	Not published	128k	64k	128k	128k	128k	Not published	128k	Not published	384k	384k	128k	128k	131.1k	Not published	65.5k	Not published	128k	128k	128k	384k	384k	Not published	128k	64k	Not published	131.1k	64k	128k	Not published	262.1k	128k
Modalities
Input	Text, Image	Text, Image	Text, Image	Text, Image	Text, Image	Text, Image, File	Text, Image	Text, Image	Text, Image	Text, Image	Text, Image	Text	Text	Text, Image, Video	Text	Text, Image, Video	Text	Text, Image	Text	Text, Image	Text, Image	Text, Image	Text	Text	Not published	Text, Image	Text, Image, Audio, Video	Text, Image, Audio, Video	Text, Image, Audio, Video	Text, Image, Video	Text, Image	Text, Image	Text, Image	Text, Image
Output	Text	Text	Text	Text	Text	Text	Text	Text	Text	Text	Text	Text	Text	Text	Text	Text	Text	Text	Text	Text	Text	Text	Text	Text	Not published	Text	Text	Text	Text	Text	Text	Text	Text	Text
Capabilities
Tools	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Vision	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	No	Yes	No	Yes	No	Yes	No	Yes	Yes	Yes	No	No	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Reasoning	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	No	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
JSON mode	No	Yes	Yes	No	No	Yes	Yes	Yes	No	No	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	No	No	Yes	Yes	No	No	No	No	No	No	No	No	No	No
Structured output	Yes	Yes	No	Yes	Yes	No	Yes	Yes	Yes	Yes	Yes	No	No	No	No	No	No	No	No	Yes	Yes	Yes	No	No	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Prompt caching	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	No	No	No	No	No	No	No	Yes	Yes	No	Yes	Yes	Yes	No	Yes	Yes
Access
API formats	Anthropic, OpenAI	OpenAI	OpenAI	Anthropic, OpenAI	Anthropic, OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	Anthropic, OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI	OpenAI
Plan allowances	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max	Pro, Plus, Max
Released	22 Sept 2026	29 Sept 2026	Not published	30 Jun 2026	15 Oct 2025	Not published	9 Jul 2026	9 Jul 2026	Not published	Not published	21 Sept 2026	10 Sept 2026	13 Aug 2026	Not published	Not published	Not published	Not published	Not published	Not published	Not published	Not published	Not published	Not published	Not published	Not published	Not published	Not published	Not published	Not published	Not published	Not published	Not published	Not published	Not published
Benchmarks reported by the vendor
Published scoresNot re-run by us; compare identical names only	Terminal-Bench 4.0 (xhigh effort)66.4% Humanity's Last Exam, with tools (max effort)67.7% OSWorld 2.1 (partial score, max effort)81.8% Source for Claude Opus 5.5 scores (opens in a new tab)	Not published	GPQA Diamond (max effort)93.5% HLE, no tools (max effort)43.5% Terminal-Bench 2.1 (Kimi Code harness, max effort)88.3% DeepSWE v1.1 (Kimi Code harness, max effort)67.5% Source for Kimi K3 scores (opens in a new tab)	Humanity's Last Exam, with tools54.9% OSWorld 2.1 (partial score)57% Source for Claude Sonnet 5 scores (opens in a new tab)	SWE-bench Verified (average of 50 trials)73.3% Source for Claude Haiku 4.5 scores (opens in a new tab)	GPQA Diamond (best effort setting)96% Terminal-Bench 4.0 (best effort setting)57.9% DeepSWE v1.1 (best effort setting)74.1% OSWorld 2.0 (offline set, partial score)72.6% Source for GPT-6 Astra scores (opens in a new tab)	GPQA Diamond94.6% Terminal-Bench 2.188.8% SWE-Bench Pro64.6% DeepSWE v1.172.7% Source for GPT-5.6 Sol scores (opens in a new tab)	GPQA Diamond92.9% Terminal-Bench 2.187.4% SWE-Bench Pro63.4% DeepSWE v1.169.6% Source for GPT-5.6 Terra scores (opens in a new tab)	Terminal-Bench 4.0 (High)20.3% DeepSWE v1.1 (High)65.2% CursorBench 4.0 (High)40.4% Source for Grok 4.6 scores (opens in a new tab)	Terminal-Bench 4.070.6% Humanity's Last Exam, with tools64.5% OSWorld 2.1 (partial score)80.1% Source for Claude Sonnet 5.5 scores (opens in a new tab)	Terminal-Bench 4.0 (xHigh)37.6% DeepSWE v1.1 (high effort)71% CursorBench 4.0 (xHigh)46.3% Source for Grok 4.7 scores (opens in a new tab)	GPQA Diamond (pass@1, reasoning effort 100)90.9% HLE (pass@1, no tools, max effort)36.8% Terminal-Bench 2.1 (pass@1, max effort)90.6% DeepSWE v1.1 (mini-SWE harness, max effort)74.2% Source for DeepSeek V4 Flash scores (opens in a new tab)	SWE-bench Verified (Max mode)80.6% GPQA Diamond (pass@1, Max mode)90.1% HLE (pass@1, Max mode)37.7% Terminal-Bench 2.0 (Max mode)67.9% Source for DeepSeek V4 Pro scores (opens in a new tab)	GPQA Diamond92.6% HLE, no tools43.6% Terminal-Bench 2.1 (Claude Code harness, avg@10)86.6% SWE-bench Pro (Claude Code harness)67.7% Source for Qwen3.8 Max scores (opens in a new tab)	Terminal-Bench 2.1 (Claude Code harness)88.2% DeepSWE v1.1 (mini-swe-agent)66.9% HLE, with tools62.5% Source for GLM 5.3 scores (opens in a new tab)	SWE-Bench Pro (Claude Code scaffold)59% Terminal-Bench 2.1 (Terminus 2 scaffold)66% MCP Atlas74.2% Source for MiniMax M3 scores (opens in a new tab)	DeepSWE v1.163.4% AutomationBench48.8% Source for GLM 5.3 Flash scores (opens in a new tab)	Not published	Not published	Not published	Not published	Not published	Not published	Not published	Not published	Humanity's Last Exam, no tools60.9% Humanity's Last Exam, with tools65% Terminal-Bench 4.055.8% OSWorld 2.0 (partial score)77.9% Source for Claude Fable 5.1 scores (opens in a new tab)	Terminal-Bench 4.057.4% DeepSWE v1.177.9% OSWorld 2.0 (offline subset, partial score)69.2% Source for Gemini 4 Argon scores (opens in a new tab)	DeepSWE v1.1 (max effort, run by Datacurve)78.8% OSWorld 2.1 (partial score, max effort)72.6% Source for Muse Wonder scores (opens in a new tab)	DeepSWE v1.171.9% Terminal-Bench 2.189.9% Terminal-Bench 4.034.9% OSWorld-Verified82% Source for MiMo V2.6 Pro scores (opens in a new tab)	Not published	Humanity's Last Exam, no tools45.9% Humanity's Last Exam, with tools57.4% OSWorld 2.1 (offline subset)72.4% Source for Claude Haiku 5.5 scores (opens in a new tab)	GPQA Diamond (xhigh)89% HLE, with tools (xhigh)61% Terminal-Bench 2.1 (Terminus 2 agent, xhigh)76.4% Source for Muse Spark Open scores (opens in a new tab)	DeepSWE v1.1 (public preview)61.7% Terminal-Bench 4 (public preview)28.3% SWE-Atlas-QnA (public preview)59.4% Source for Mistral Large 4 scores (opens in a new tab)	DeepSWE v1.1 (max effort)66.6% Source for GPT-6 Luna scores (opens in a new tab)
Links	Model page for Claude Opus 5.5 Try in Chat with Claude Opus 5.5	Model page for GPT-6.1 Sol Try in Chat with GPT-6.1 Sol	Model page for Kimi K3 Try in Chat with Kimi K3	Model page for Claude Sonnet 5 Try in Chat with Claude Sonnet 5	Model page for Claude Haiku 4.5 Try in Chat with Claude Haiku 4.5	Model page for GPT-6 Astra Try in Chat with GPT-6 Astra	Model page for GPT-5.6 Sol Try in Chat with GPT-5.6 Sol	Model page for GPT-5.6 Terra Try in Chat with GPT-5.6 Terra	Model page for Grok 4.6 Try in Chat with Grok 4.6	Model page for Claude Sonnet 5.5 Try in Chat with Claude Sonnet 5.5	Model page for Grok 4.7 Try in Chat with Grok 4.7	Model page for DeepSeek V4 Flash Try in Chat with DeepSeek V4 Flash	Model page for DeepSeek V4 Pro Try in Chat with DeepSeek V4 Pro	Model page for Qwen3.8 Max Try in Chat with Qwen3.8 Max	Model page for GLM 5.3 Try in Chat with GLM 5.3	Model page for MiniMax M3 Try in Chat with MiniMax M3	Model page for GLM 5.3 Flash Try in Chat with GLM 5.3 Flash	Model page for Qwen3.8 27B Try in Chat with Qwen3.8 27B	Model page for Qwen3.8 27B Groq Try in Chat with Qwen3.8 27B Groq	Model page for Claude Fable 5.1 (free lane)Try in Chat with Claude Fable 5.1 (free lane)	Model page for Claude Opus 5.5 (free lane)Try in Chat with Claude Opus 5.5 (free lane)	Model page for Claude Sonnet 5.5 (free lane)Try in Chat with Claude Sonnet 5.5 (free lane)	Model page for DeepSeek V4 Flash (free lane)Try in Chat with DeepSeek V4 Flash (free lane)	Model page for DeepSeek V4 Pro (free lane)Try in Chat with DeepSeek V4 Pro (free lane)	Model page for StepAudio 3 Chat Preview Try in Chat with StepAudio 3 Chat Preview	Model page for Claude Fable 5.1 Try in Chat with Claude Fable 5.1	Model page for Gemini 4 Argon Try in Chat with Gemini 4 Argon	Model page for Muse Wonder Try in Chat with Muse Wonder	Model page for MiMo V2.6 Pro Try in Chat with MiMo V2.6 Pro	Model page for Step 5 Preview Try in Chat with Step 5 Preview	Model page for Claude Haiku 5.5 Try in Chat with Claude Haiku 5.5	Model page for Muse Spark Open Try in Chat with Muse Spark Open	Model page for Mistral Large 4 Try in Chat with Mistral Large 4	Model page for GPT-6 Luna Try in Chat with GPT-6 Luna

Rates in US dollars per 1M tokens; struck-through rates are the lowest public price for the same model at other API gateways. Example costs use today's rates and include the instructions a model adds to every request itself. Model your own traffic

Switching later is one line.

Every model here runs on the same key and balance, so you can start with one and change the model field when the work changes.

Get an API key All models

Two or three models, side by side.

Tokens per day, by model

Comparison

Switching later is one line.