Context windows

Each label is the context window printed on that model’s page. 256k, 128k, 1M, 1024k, and 16k are left as printed. This page does not turn them into a token integer the source did not state. Gemini, Cohere, Amazon Bedrock, and Azure OpenAI have no row: a context integer was not copied from their docs in this pass. Cache and long-context pricing are in the notes, not in the window column.

Filter rows

Provider	Model	Context as printed	Note	Source
OpenAI	gpt-6-luna	1,050,000 tokens	Model page also lists cached input at $0.01 / 1M and cache writes at $0.125 / 1M. Prompts with more than 272K input tokens are priced at 2× input and cache rates and 1.5× output for the full request. Batch and Flex are 50% of Standard. Fast mode is 2×.	Official page
OpenAI	gpt-6.1-sol	1,050,000 tokens	Cached input $0.10 / 1M. Cache writes $2.50 / 1M. The same >272K full-request multiplier applies. Fast mode is 2× Standard. Batch and Flex are 50% lower than Standard. Ultrafast mode is 6× Standard. Regional processing adds a 10% premium where available.	Official page
Anthropic	claude-opus-5-5	1M tokens	Models overview lists this as the standard base price and a 1M-token context window, with 128K max output. The pricing page lists separate cache-write and cache-hit rates.	Official page
Anthropic	claude-sonnet-5-5	1M tokens	Models overview lists $2 / input MTok and $10 / output MTok, a 1M-token context window, and 128K max output.	Official page
Anthropic	claude-haiku-5-5 (prompt up to 100,000 tokens)	1M tokens	Pricing page rate for prompts up to 100,000 tokens. The models overview prints the price as “From $0.10 / input MTok” and “From $0.50 / output MTok” and the context window as 1M tokens. A request whose prompt is over 100,000 tokens uses the higher row.	Official page
Anthropic	claude-haiku-5-5 (prompt over 100,000 tokens)	1M tokens	Pricing page: prompts over 100,000 tokens. Prompt length counts input tokens, including cache reads and cache writes.	Official page
Groq	openai/gpt-oss-120b	131,072 tokens	Production-models row on the Groq models page. Developer-plan limits on the same row: 250K TPM and 1K RPM. Max completion tokens: 65,536. These are listed prices, not the Free Plan Limits table.	Official page
Groq	openai/gpt-oss-20b	131,072 tokens	Production-models row. Developer-plan limits on the same row: 250K TPM and 1K RPM. Max completion tokens: 65,536.	Official page
Cerebras	gpt-oss-120b	65k tokens (free tier), 131k tokens (paid)	Model card developer pricing. The same card states free-tier rate limits of 30 requests/min, 60k input tokens/min, and 1M tokens/day. The rate-limits page lists a different Free Trial row for this id (5 RPM, 30K uncached TPM). Both pages are linked from the Cerebras free-tier note.	Official page
Cerebras	qwen-3.8-27b	65,536 tokens on Free Trial; 131,072 on paid tiers	Model card. Free Trial rate limits on that card: 5 requests per minute, 30K uncached tokens per minute, 90K total tokens per minute, and 1M tokens per day. Those request limits match the rate-limits page.	Official page
Mistral	mistral-small-2603	256k	Model card also lists the alias mistral-small-latest. The card prints context length as 256k, not an integer token count. Features named on the card include function-calling.	Official page
Mistral	codestral-2508	128k	Model card also lists the alias codestral-latest. Context length is printed as 128k.	Official page
Together	zai-org/GLM-5.3-Flash	1M	Model page: input $0.15 / 1M tokens, cached input $0.03 / 1M, output $0.50 / 1M tokens. Context window label is 1M.	Official page
Together	moonshotai/Kimi-K3	1M	Model page: input $3.00 / 1M tokens, cached input $0.30 / 1M, output $15.00 / 1M tokens. Context window label is 1M.	Official page
DeepInfra	DeepSeek-V4-Flash-0731	1024k	Pricing table writes the input cell as “$0.06 / $0.015 cached” and the context cell as 1024k. Href on that row: /deepseek-ai/DeepSeek-V4-Flash-0731.	Official page
DeepInfra	phi-4	16k	Pricing table row. Href on that row: /microsoft/phi-4. No cached price is printed in the input cell.	Official page
Cloudflare Workers AI	@cf/meta/llama-3.2-1b-instruct	60,000 tokens	Model page unit pricing. The platform pricing page lists the same model at 2457 neurons per million input tokens and 18252 neurons per million output tokens. Workers AI also has a separate daily allocation of 10,000 Neurons.	Official page
Cloudflare Workers AI	@cf/deepseek-ai/deepseek-v4.1-flash-1017	1,048,576 tokens	Model page unit pricing also lists cached input at $0.006 per million tokens. The page says paid access is required. The platform pricing page lists this id among models that require a paid billing method.	Official page

Sources

Last checked 2026-10-22.