{"schema_version":"1.0","slug":"llm-api-pricing-2026","title":"LLM API Pricing 2026: Full Comparison Table (Weekly)","description":"LLM API pricing 2026: 54 models from OpenAI, Anthropic, Google, xAI, DeepSeek, and Mistral in one table. Input/output cost per 1M tokens, updated weekly.","data_updated":"2026-10-26","source_post":"https://www.jsonhouse.com/posts/llm-api-pricing-2026/","category":"ai-data-statistics","cluster":"CLUSTER_LLM","format":"D","attribution":{"source":"Json House","source_url":"https://www.jsonhouse.com/posts/llm-api-pricing-2026/","dataset_url":"https://www.jsonhouse.com/data/llm-api-pricing-2026.json","citation":"Json House, \"LLM API Pricing 2026: Full Comparison Table (Weekly)\", jsonhouse.com (2026-10-26)","attribution_required":true,"terms_url":"https://www.jsonhouse.com/data-policy/"},"key_facts":[{"fact":"DeepSeek retired v4-flash on 2026-09-14 and serves the same model name from deepseek-flash (build DeepSeek-V4.1-Flash) at $0.15/$0.60 per 1M tokens off-peak; against the 2026-08-17 increase to $0.22/$0.66 this reverses the rise on input (+7% versus the pre-increase $0.14) but not on output (+114% versus the pre-increase $0.28), taking DeepSeek output-to-input from 2x to 3x to 4x across the three states. On 2026-10-17 DeepSeek replaced the build behind deepseek-flash with DeepSeek-V4.1-Flash-1017 at unchanged prices, a second substitution under the same id","source":"DeepSeek pricing page, retrieved 2026-09-21 and 2026-10-19; DeepSeek API change log entry of 2026-10-17","category":"data"},{"fact":"Anthropic cancelled the scheduled 2026-09-01 increase of Claude Sonnet 5 from $2/$10 to $3/$15 per 1M tokens; the introductory rate became the standard price","source":"Anthropic pricing page, retrieved 2026-08-17","category":"data"},{"fact":"No price carried into the tracked set changed in the six collections from 2026-09-01 to 2026-10-05; the run ended on 2026-10-14 with a cut, Claude Sonnet 5.5's cache read falling from $0.20 to $0.10 per 1M tokens — the first carried-row price change since 2026-08-24 and the first reduction since then to reach an existing model id. Input and output prices on all 51 carried rows held. The 2026-10-19 collection held all 54 rows on input, output and cache read. On 2026-10-26 Mistral Large 4 rose 100% on input, output and cache read ($0.68/$2.09/$0.07 to $1.36/$4.18/$0.14 per 1M tokens) when an undated sale ended, the first carried-row increase since DeepSeek's on 2026-08-17; the other 53 rows held","source":"jsonhouse weekly pricing snapshot series, collections of 2026-09-01 through 2026-10-26; Mistral API pricing page, retrieved 2026-10-26","category":"data"},{"fact":"Claude 4.7 and later models, plus Claude Mythos Preview, use a tokenizer that produces approximately 30% more tokens for the same text than Sonnet 4.6 and earlier — 11 of the 16 tracked Anthropic rows as of 2026-10-14 — so an unchanged per-token price on those rows is not an unchanged cost per unit of text, and per-token comparison across that line or across providers is an approximation","source":"Anthropic pricing page, retrieved 2026-10-14","category":"definition"},{"fact":"Anthropic now runs three cache-read multipliers where it ran one before 2026-09-07: 0.025x base input on Claude Fable 5.1 and Mythos 5.1 ($0.25 per 1M, 97.5% off), 0.05x on Claude Opus 5.5 from 2026-09-28 ($0.20, 95% off) and on Claude Sonnet 5.5 from 2026-10-14 ($0.10), and 0.1x on every other Claude model; DeepSeek holds the deepest discount overall at 98% ($0.003 off a $0.15 base). OpenAI left the flat 90% on 2026-10-05 with gpt-6.1-sol at 0.05x ($0.10 off a $2.00 base), and Google followed on 2026-10-14 with Gemini 4 Argon at the same 0.05x, so none of the three original 0.1x vendors is still flat across its line","source":"Official pricing pages of Anthropic, OpenAI, Google, DeepSeek, retrieved 2026-10-14","category":"data"},{"fact":"Long-context pricing structures diverged: Google and xAI bill roughly 2x per token above 200K-token prompts, OpenAI added long-context tiers from 2026-08-24 (GPT-5.6-sol $8.00/$30.00 against a $4.00/$20.00 base), and Anthropic bills its full 1M-token window at standard rates on every Claude 4.6-and-later model except Claude Haiku 5.5, added 2026-10-14, which charges $0.50/$2.50 above 100,000 prompt tokens against a $0.10/$0.50 base","source":"Google, xAI, OpenAI, Anthropic pricing pages, retrieved 2026-10-14","category":"trend"},{"fact":"Three of the four models that entered the tracked set on 2026-09-28 are priced below the generation they succeed — Claude Opus 5.5 at $4.00/$20.00 against Claude Opus 5's $5.00/$25.00, and gpt-6-sol ($2.00/$10.00) and gpt-6-luna ($0.10/$0.50) at exactly half of gpt-5.6-sol and gpt-5.6-luna on both sides — so the frontier price fell in the same week that no carried row moved; the saving reaches only a workload that changes model id","source":"jsonhouse weekly pricing snapshot series, 2026-09-28 collection against vendor pricing pages","category":"data"},{"fact":"Both models that entered the tracked set on 2026-10-05 carry exactly the input and output price of the model they succeed — Claude Sonnet 5.5 matches Claude Sonnet 5 at $2.00/$10.00 on every line of Anthropic's table, and gpt-6.1-sol matches gpt-6-sol at $2.00/$10.00 — but gpt-6.1-sol bills a cache read at $0.10 against gpt-6-sol's $0.20, the first OpenAI row below a 0.1x cache-read multiplier; the reduction is invisible to a list-price comparison","source":"jsonhouse weekly pricing snapshot series, 2026-10-05 collection against Anthropic and OpenAI pricing pages","category":"data"},{"fact":"OpenAI cut GPT-5.6-sol 20% on input and 33% on output to $4.00 / $20.00 on 2026-08-24, a promotional rate stated as available at least through 2026-11-21; earlier cuts on 2026-08-03 took GPT-5.6-luna down 80% to $0.20 / $1.20 and GPT-5.6-terra down 20% to $2.00 / $12.00, neither of which moved this week","source":"OpenAI API pricing page, retrieved 2026-08-24","category":"data"},{"fact":"Three models entered the tracked set on 2026-10-14 and two left: Claude Haiku 5.5 at $0.10/$0.50 per 1M tokens (one tenth of Claude Haiku 4.5, tying gpt-6-luna and Gemini 2.5 Flash-Lite on input and cache read), Gemini 4 Argon at $2.00/$10.00 introductory through 2026-11-30 and $4.00/$20.00 from 2026-12-01, and Mistral Large 4 at a sale price of $0.68/$2.09 against a list price of $1.36/$4.18; Google removed Gemini 3.7 Flash and 3.5 Flash from its pricing page and routes their ids to Gemini 3.8 Flash and 3.6 Flash without stating how routed requests are billed","source":"jsonhouse weekly pricing snapshot series, 2026-10-14 collection against Anthropic, Google and Mistral pricing pages and the Gemini API deprecations page","category":"data"}],"faq_summary":[{"q":"What is the cheapest LLM API in 2026?","a":"Gemini 2.5 Flash-Lite at $0.10 input / $0.40 output per 1M tokens, though it shares that floor: gpt-6-luna (from 2026-09-28) and Claude Haiku 5.5 (from 2026-10-14) both charge $0.10 / $0.50, matching it on input and on the $0.01 cache read and separating only on output. DeepSeek is next at $0.15 / $0.60, tied with Mistral Small 4, after retiring v4-flash on 2026-09-14 and serving the name from deepseek-flash. DeepSeek leads on cached input at $0.003 per 1M tokens, a 98% discount."},{"q":"Why is Claude Fable 5.1's cache read cheaper than Claude Opus 5's?","a":"Anthropic prices a cache hit as a multiplier on base input and stopped applying one multiplier to the whole line on 2026-09-07. There are now three. Fable 5.1 and Mythos 5.1 bill a cache hit at 0.025x base input ($0.25 per 1M); Claude Opus 5.5 bills at 0.05x ($0.20), and Claude Sonnet 5.5 joined it at 0.05x ($0.10) on 2026-10-14; every other Claude model, Opus 5 and Haiku 5.5 included, stays at 0.1x. The deeper multipliers now track recency more than price."},{"q":"Are per-token prices directly comparable across providers?","a":"Only approximately. Anthropic's newest models tokenize the same text into roughly 30% more tokens than earlier Claude models, so identical sticker prices can produce different invoices. Compare cost per task, not cost per token."},{"q":"Do these prices include long-context surcharges?","a":"No — base tiers only. Google and xAI roughly double rates above 200K-token prompts; Anthropic bills its 1M context at standard rates except on Claude Haiku 5.5, which charges 5x above 100,000 prompt tokens."},{"q":"How often is this table updated?","a":"Weekly, from official provider pages only, with every change recorded in the changelog. Missed weeks are marked as gaps rather than back-filled: the 2026-07-20 and 2026-07-27 snapshots were not taken, and the 2026-10-12 collection was taken on 2026-10-14 and dated as such. The series began 2026-07-16."}],"primary_sources":[{"title":"Anthropic API pricing","url":"https://platform.claude.com/docs/en/about-claude/pricing","publisher":"Anthropic"},{"title":"OpenAI API pricing","url":"https://developers.openai.com/api/docs/pricing","publisher":"OpenAI"},{"title":"Gemini API pricing","url":"https://ai.google.dev/gemini-api/docs/pricing","publisher":"Google"},{"title":"xAI models and pricing","url":"https://docs.x.ai/docs/models","publisher":"xAI"},{"title":"DeepSeek API pricing","url":"https://api-docs.deepseek.com/quick_start/pricing","publisher":"DeepSeek"},{"title":"Mistral API pricing","url":"https://mistral.ai/pricing/api","publisher":"Mistral AI"}],"comparison_data":{"field_provenance":{"input_per_1m_usd":"vendor","output_per_1m_usd":"vendor","cache_read_per_1m_usd":"vendor","context_window":"vendor","tier":"jsonhouse editorial grouping","note":"vendor, except dated change descriptions which are jsonhouse observations from the weekly snapshot series"},"dimensions":["provider","tier","input_per_1m_usd","output_per_1m_usd","cache_read_per_1m_usd","context_window","note"],"unit":"USD per 1M tokens, standard non-batch tier, collected 2026-10-26","entries":[{"model":"Claude Fable 5.1","provider":"Anthropic","tier":"flagship","input_per_1m_usd":10.0,"output_per_1m_usd":50.0,"cache_read_per_1m_usd":0.25,"context_window":1000000,"note":"Added 2026-09-07; cache hits billed at 0.025x base input against the 0.1x every other Claude model uses. Claude Mythos 5.1 carries identical pricing but is application-gated"},{"model":"Claude Fable 5","provider":"Anthropic","tier":"flagship","input_per_1m_usd":10.0,"output_per_1m_usd":50.0,"cache_read_per_1m_usd":1.0,"context_window":1000000,"note":null},{"model":"GPT-6-astra","provider":"OpenAI","tier":"flagship","input_per_1m_usd":10.0,"output_per_1m_usd":50.0,"cache_read_per_1m_usd":1.0,"context_window":null,"note":"Added 2026-09-07; $20.00/$75.00 above the base context window (cache read $2.00). Most expensive OpenAI row outside the $30/$180 pro tier"},{"model":"GPT-5.5-pro","provider":"OpenAI","tier":"flagship","input_per_1m_usd":30.0,"output_per_1m_usd":180.0,"cache_read_per_1m_usd":null,"context_window":null,"note":null},{"model":"GPT-5.4-pro","provider":"OpenAI","tier":"flagship","input_per_1m_usd":30.0,"output_per_1m_usd":180.0,"cache_read_per_1m_usd":null,"context_window":null,"note":null},{"model":"Claude Opus 5.5","provider":"Anthropic","tier":"flagship","input_per_1m_usd":4.0,"output_per_1m_usd":20.0,"cache_read_per_1m_usd":0.2,"context_window":1000000,"note":"Added 2026-09-28 at $1.00/$5.00 below the Claude Opus 5 it succeeds; cache hits billed at 0.05x base input, a third Anthropic multiplier between Fable 5.1's 0.025x and the 0.1x the rest of the line carries. Fast mode $8.00/$40.00"},{"model":"GPT-5.6-sol","provider":"OpenAI","tier":"flagship","input_per_1m_usd":4.0,"output_per_1m_usd":20.0,"cache_read_per_1m_usd":0.4,"context_window":null,"note":"Cut from $5.00/$30.00 on 2026-08-24 (cache read $0.50 to $0.40); promotional, stated available at least through 2026-11-21. Long-context tier $8.00/$30.00, cache read $0.80 — OpenAI's first tiered row"},{"model":"GPT-5.5","provider":"OpenAI","tier":"flagship","input_per_1m_usd":5.0,"output_per_1m_usd":30.0,"cache_read_per_1m_usd":0.5,"context_window":null,"note":null},{"model":"Claude Opus 5","provider":"Anthropic","tier":"flagship","input_per_1m_usd":5.0,"output_per_1m_usd":25.0,"cache_read_per_1m_usd":0.5,"context_window":1000000,"note":"Claude Opus 4.5 through 4.8 share this price"},{"model":"Claude Opus 4.8","provider":"Anthropic","tier":"flagship","input_per_1m_usd":5.0,"output_per_1m_usd":25.0,"cache_read_per_1m_usd":0.5,"context_window":1000000,"note":"Opus 4.5, 4.6, and 4.7 share this pricing"},{"model":"GPT-6.1-sol","provider":"OpenAI","tier":"flagship","input_per_1m_usd":2.0,"output_per_1m_usd":10.0,"cache_read_per_1m_usd":0.1,"context_window":null,"note":"Added 2026-10-05 at GPT-6-sol's list price with the cache read halved to 0.05x base input, the first OpenAI row below 0.1x. Long-context tier $4.00/$15.00, cache read $0.20. Replaced GPT-6-sol in OpenAI's three-row flagship table; GPT-6-sol remains listed"},{"model":"GPT-6-sol","provider":"OpenAI","tier":"flagship","input_per_1m_usd":2.0,"output_per_1m_usd":10.0,"cache_read_per_1m_usd":0.2,"context_window":null,"note":"Added 2026-09-28 at exactly half of GPT-5.6-sol on input and output, and on the same published rate as Claude Sonnet 5. Long-context tier $4.00/$15.00, cache read $0.40"},{"model":"Gemini 4 Argon","provider":"Google","tier":"flagship","input_per_1m_usd":2.0,"output_per_1m_usd":10.0,"cache_read_per_1m_usd":0.1,"context_window":1048576,"note":"Added 2026-10-14 (GA 2026-10-13); introductory through 2026-11-30, then $4.00/$20.00 (cache read $0.20) from 2026-12-01. Implicit caching only; no free tier. Matches GPT-6.1-sol on all three figures"},{"model":"Gemini 3.1 Pro Preview","provider":"Google","tier":"flagship","input_per_1m_usd":2.0,"output_per_1m_usd":12.0,"cache_read_per_1m_usd":0.2,"context_window":null,"note":"$4.00/$18.00 above 200K-token prompts"},{"model":"Grok 4.7","provider":"xAI","tier":"flagship","input_per_1m_usd":2.0,"output_per_1m_usd":6.0,"cache_read_per_1m_usd":0.5,"context_window":500000,"note":"Added 2026-09-28 on Grok 4.6's rates to the cent, cache read included; billed $4.00 / $12.00 above 200K tokens (cache read $1.00)"},{"model":"Grok 4.6","provider":"xAI","tier":"flagship","input_per_1m_usd":2.0,"output_per_1m_usd":6.0,"cache_read_per_1m_usd":0.5,"context_window":500000,"note":"Added 2026-08-17; billed $4.00 / $12.00 above 200K tokens (cache read $1.00); matches Grok 4.5 on base rate but costs more on cache read"},{"model":"Grok 4.5","provider":"xAI","tier":"flagship","input_per_1m_usd":2.0,"output_per_1m_usd":6.0,"cache_read_per_1m_usd":0.3,"context_window":500000,"note":"Billed $4.00 / $12.00 above 200K tokens; cache read cut from $0.50 on 2026-07-28"},{"model":"Claude Sonnet 5.5","provider":"Anthropic","tier":"mid-range","input_per_1m_usd":2.0,"output_per_1m_usd":10.0,"cache_read_per_1m_usd":0.1,"context_window":1000000,"note":"Added 2026-10-05 at Claude Sonnet 5's rate on every line of Anthropic's table; cache read cut from $0.20 (0.1x) to $0.10 (0.05x) between 2026-10-05 and 2026-10-14, the first carried-row price change since 2026-08-24. Newer Claude 4.7-and-later tokenizer"},{"model":"Claude Sonnet 5","provider":"Anthropic","tier":"mid-range","input_per_1m_usd":2.0,"output_per_1m_usd":10.0,"cache_read_per_1m_usd":0.2,"context_window":1000000,"note":"Introductory rate became the standard price on 2026-08-17; the scheduled 2026-09-01 rise to $3.00/$15.00 was cancelled"},{"model":"Claude Sonnet 4.6","provider":"Anthropic","tier":"mid-range","input_per_1m_usd":3.0,"output_per_1m_usd":15.0,"cache_read_per_1m_usd":0.3,"context_window":1000000,"note":null},{"model":"GPT-5.6-terra","provider":"OpenAI","tier":"mid-range","input_per_1m_usd":2.0,"output_per_1m_usd":12.0,"cache_read_per_1m_usd":0.2,"context_window":null,"note":"Cut from $2.50 / $15.00 (cache read $0.25) on 2026-08-03"},{"model":"GPT-5.4","provider":"OpenAI","tier":"mid-range","input_per_1m_usd":2.5,"output_per_1m_usd":15.0,"cache_read_per_1m_usd":0.25,"context_window":null,"note":null},{"model":"Gemini 3.8 Flash","provider":"Google","tier":"mid-range","input_per_1m_usd":0.75,"output_per_1m_usd":3.75,"cache_read_per_1m_usd":0.075,"context_window":null,"note":"Added 2026-09-07 at the rate Gemini 3.7 and 3.6 Flash already carried; promotional through 2026-12-31, doubles to $1.50/$7.50 (cache read $0.15) on 2027-01-01. Since 2026-10-08 Google routes requests for the deprecated gemini-3.7-flash id to this model"},{"model":"Gemini 3.6 Flash","provider":"Google","tier":"mid-range","input_per_1m_usd":0.75,"output_per_1m_usd":3.75,"cache_read_per_1m_usd":0.075,"context_window":null,"note":"Halved from $1.50/$7.50 on 2026-08-17 to match the Gemini 3.7 Flash launch rate; same 2026-12-31 promotional expiry. Since 2026-10-08 Google routes requests for the deprecated gemini-3.5-flash id ($1.50/$9.00 when last listed) to this model"},{"model":"Mistral Medium 3.5","provider":"Mistral","tier":"mid-range","input_per_1m_usd":1.5,"output_per_1m_usd":7.5,"cache_read_per_1m_usd":0.15,"context_window":null,"note":"Cached input first recorded 2026-10-14"},{"model":"Grok 4.3","provider":"xAI","tier":"mid-range","input_per_1m_usd":1.25,"output_per_1m_usd":2.5,"cache_read_per_1m_usd":0.2,"context_window":1000000,"note":"$2.50/$5.00 above 200K-token prompts"},{"model":"DeepSeek v4-pro","provider":"DeepSeek","tier":"mid-range","input_per_1m_usd":0.66,"output_per_1m_usd":1.98,"cache_read_per_1m_usd":0.022,"context_window":1000000,"note":"Off-peak rate; peak is exactly double ($1.32/$3.96, cache read $0.044) during 01:00-04:00 and 06:00-10:00 UTC on weekdays; weekends and Chinese public holidays are off-peak. Raised from $0.435/$0.87 on 2026-08-17; build V4-Pro-0813"},{"model":"Claude Haiku 4.5","provider":"Anthropic","tier":"budget","input_per_1m_usd":1.0,"output_per_1m_usd":5.0,"cache_read_per_1m_usd":0.1,"context_window":200000,"note":null},{"model":"GPT-5.6-luna","provider":"OpenAI","tier":"budget","input_per_1m_usd":0.2,"output_per_1m_usd":1.2,"cache_read_per_1m_usd":0.02,"context_window":null,"note":"Cut from $1.00 / $6.00 (cache read $0.10) on 2026-08-03; now matches GPT-5.4-nano on input and cache read"},{"model":"GPT-5.4-mini","provider":"OpenAI","tier":"budget","input_per_1m_usd":0.75,"output_per_1m_usd":4.5,"cache_read_per_1m_usd":0.075,"context_window":null,"note":null},{"model":"Mistral Large 4","provider":"Mistral","tier":"mid-range","input_per_1m_usd":1.36,"output_per_1m_usd":4.18,"cache_read_per_1m_usd":0.14,"context_window":null,"note":"Added 2026-10-14 at a sale price of $0.68/$2.09 (cached $0.07) beside this list price, with no end date stated; the sale ended between 2026-10-19 and 2026-10-26, a 100% increase on all three prices, and the row moved from the budget tier to mid-range"},{"model":"Mistral Large 3","provider":"Mistral","tier":"budget","input_per_1m_usd":0.5,"output_per_1m_usd":1.5,"cache_read_per_1m_usd":0.05,"context_window":null,"note":"Former flagship, repositioned as budget option. Cached input first recorded 2026-10-14"},{"model":"Gemini 2.5 Flash","provider":"Google","tier":"budget","input_per_1m_usd":0.3,"output_per_1m_usd":2.5,"cache_read_per_1m_usd":0.03,"context_window":1000000,"note":null},{"model":"Gemini 3.1 Flash-Lite","provider":"Google","tier":"budget","input_per_1m_usd":0.25,"output_per_1m_usd":1.5,"cache_read_per_1m_usd":0.025,"context_window":null,"note":null},{"model":"GPT-5.4-nano","provider":"OpenAI","tier":"budget","input_per_1m_usd":0.2,"output_per_1m_usd":1.25,"cache_read_per_1m_usd":0.02,"context_window":null,"note":"Deprecated: removal from the API scheduled for 2027-04-01, replacement gpt-6-luna, per an OpenAI deprecations entry dated 2026-10-01; first recorded in this series on 2026-10-26"},{"model":"Mistral Small 4","provider":"Mistral","tier":"budget","input_per_1m_usd":0.15,"output_per_1m_usd":0.6,"cache_read_per_1m_usd":0.015,"context_window":null,"note":"Cached input first recorded 2026-10-14"},{"model":"Claude Haiku 5.5","provider":"Anthropic","tier":"budget","input_per_1m_usd":0.1,"output_per_1m_usd":0.5,"cache_read_per_1m_usd":0.01,"context_window":1000000,"note":"Added 2026-10-14 at one tenth of Claude Haiku 4.5. Rate applies to prompts up to 100,000 tokens; above that $0.50/$2.50, cache read $0.05, with cache reads and writes counted toward the threshold. Newer Claude 4.7-and-later tokenizer"},{"model":"DeepSeek flash","provider":"DeepSeek","tier":"budget","input_per_1m_usd":0.15,"output_per_1m_usd":0.6,"cache_read_per_1m_usd":0.003,"context_window":1000000,"note":"Off-peak rate; peak is exactly double ($0.30/$1.20, cache read $0.006) during 01:00-04:00 and 06:00-10:00 UTC on weekdays; weekends and Chinese public holidays are off-peak. Replaces the retired v4-flash (build V4-Flash-0731) on 2026-09-14; build DeepSeek-V4.1-Flash, replaced on 2026-10-17 by DeepSeek-V4.1-Flash-1017 at the same price. The legacy id deepseek-v4-flash is still accepted and served by this model at this price, so the change is a substitution, not a price cut"},{"model":"GPT-6-luna","provider":"OpenAI","tier":"budget","input_per_1m_usd":0.1,"output_per_1m_usd":0.5,"cache_read_per_1m_usd":0.01,"context_window":null,"note":"Added 2026-09-28 at half of GPT-5.6-luna on input, output and cache read; ties Gemini 2.5 Flash-Lite on input and cache read, separating only on output. Long-context tier $0.20/$0.75, cache read $0.02"},{"model":"Gemini 2.5 Flash-Lite","provider":"Google","tier":"budget","input_per_1m_usd":0.1,"output_per_1m_usd":0.4,"cache_read_per_1m_usd":0.01,"context_window":null,"note":null},{"model":"Grok-build-0.1","provider":"xAI","tier":"specialist","input_per_1m_usd":1.0,"output_per_1m_usd":2.0,"cache_read_per_1m_usd":0.2,"context_window":256000,"note":"Coding model; $2.00/$4.00 above 200K-token prompts"},{"model":"Codestral","provider":"Mistral","tier":"specialist","input_per_1m_usd":0.3,"output_per_1m_usd":0.9,"cache_read_per_1m_usd":0.03,"context_window":null,"note":"Coding model. Cached input first recorded 2026-10-14"}]}}