25 models tracked · verified September 10, 2026
LLM API PRICING, VERIFIED AND DATED

Compare every AI model's price — before you build on it.

Input and output pricing for 10 providers — OpenAI, Anthropic, Google, Mistral, DeepSeek, xAI and more — checked directly against each provider's own pricing page, with the source and date on every row.

TRACKED MODELS · AVG $/1M TOKENS
$2.03
avg. input
$9.73
avg. output
Computed across all 25 tracked models, first verified on September 10, 2026. We re-check every 48 hours, so a trend line will appear after the next pass.
Cheapest tracked
Mistral Small 4
$0.15 / $0.60
Priciest tracked
Claude Fable 5.1 (tied)
$10.00 / $50.00
Price spread
66.7x
priciest vs. cheapest input rate
Providers tracked
10
25 models total
Input price, every tracked model
$/1M tokens · log scale · full comparison →
Claude Fable 5.1
$10.00
GPT-6 Astra
$10.00
Claude Opus 5
$5.00
GPT-5.6 Sol
$4.00
Command A
$2.50
Claude Sonnet 5
$2.00
GPT-5.6 Terra
$2.00
Gemini 3.1 Pro Preview
$2.00
Grok 4.6
$2.00
Qwen3.8-Max
$2.00
Mistral Medium 3.5
$1.50
Grok 4.3
$1.25
Amazon Nova 2 Pro
$1.25
Claude Haiku 4.5
$1.00
Grok Build 0.1
$1.00
Gemini 3.8 Flash
$0.75
Mistral Large 3
$0.50
Qwen3.7-Plus
$0.40
DeepSeek V4.1 Flash
$0.30
Amazon Nova 2 Lite
$0.30
Llama 4 Maverick
$0.27
Gemini 3.1 Flash-Lite
$0.25
GPT-5.6 Luna
$0.20
Llama 4 Scout
$0.18
Mistral Small 4
$0.15
Flagship Balanced Budget

Today's model pricing

Sources: each provider's public pricing page · verified September 10, 2026
25 of 25 models shown
Model Provider Input /1M Output /1M Context Tier Source
Mistral $0.15 $0.60 Not published Budget mistral.ai — API pricing ↗
Meta doesn't sell first-party API access — price shown is Together AI's hosted rate.
Meta (via Together AI) $0.18 $0.59 1M tokens Budget together.ai — Llama 4 Scout ↗
OpenAI $0.20 $1.20 ~1.05M tokens Budget platform.openai.com/docs — Pricing ↗
Audio input priced separately at $0.50/1M.
Google $0.25 $1.50 Not published Budget ai.google.dev — Gemini API pricing ↗
Meta doesn't sell first-party API access — price shown is Together AI's hosted rate, one of several inference providers serving this open-weight model.
Meta (via Together AI) $0.27 $0.85 ~1.05M tokens Budget together.ai — Llama 4 Maverick ↗
Peak-hour, cache-miss rate shown. Off-peak (01:00–04:00 & 06:00–10:00 UTC, Mon–Fri) is half price; cache-hit input is far cheaper still ($0.006/1M peak).
DeepSeek $0.30 $1.20 1M tokens Budget api-docs.deepseek.com — Pricing ↗
Bedrock Standard tier. Flex tier: $0.15 / $1.25.
Amazon $0.30 $2.50 1M tokens Budget aws.amazon.com — Bedrock pricing ↗
Rate shown for 0–256K input tokens; both 256K–1M tier and above bill higher on input only.
Alibaba $0.40 $1.60 1M tokens Budget alibabacloud.com — Model Studio pricing ↗
Despite the "Large" name, priced below Medium 3.5 in Mistral's current lineup.
Mistral $0.50 $1.50 Not published Budget mistral.ai — API pricing ↗
Promotional rate through Dec 31, 2026; steps up to $1.50 / $7.50 on Jan 1, 2027.
Google $0.75 $3.75 Not published Balanced ai.google.dev — Gemini API pricing ↗
Anthropic $1.00 $5.00 200K tokens Budget platform.claude.com/docs — Pricing ↗
xAI $1.00 $2.00 256K tokens Budget docs.x.ai — Pricing ↗
Long-context tier (>200K input) doubles to $2.50 / $5.
xAI $1.25 $2.50 1M tokens Balanced docs.x.ai — Pricing ↗
Preview model, Bedrock Standard tier. A cheaper "Flex" tier and pricier "Priority" tier also exist.
Amazon $1.25 $10.00 Not published (preview) Balanced aws.amazon.com — Bedrock pricing ↗
Positioned by Mistral as its top general-purpose / agentic model — priced above "Large 3".
Mistral $1.50 $7.50 Not published Balanced mistral.ai — API pricing ↗
Launched at introductory pricing through Aug 31, 2026; the scheduled increase to $3/$15 was cancelled — $2/$10 is now the standing price.
Anthropic $2.00 $10.00 1M tokens Balanced platform.claude.com/docs — Pricing ↗
OpenAI $2.00 $12.00 ~1.05M tokens Balanced platform.openai.com/docs — Pricing ↗
Requests over 200K input tokens bill at $4 / $18 instead.
Google $2.00 $12.00 ≤200K tokens tier Flagship ai.google.dev — Gemini API pricing ↗
Long-context tier (>200K input) doubles to $4 / $12.
xAI $2.00 $6.00 500K tokens Flagship docs.x.ai — Pricing ↗
Singapore-region rate, flat regardless of prompt length.
Alibaba $2.00 $6.00 1M tokens Balanced alibabacloud.com — Model Studio pricing ↗
Current flagship; priced on Cohere's docs site rather than its main pricing page.
Cohere $2.50 $10.00 256K tokens Balanced docs.cohere.com — Command A ↗
Pricing is promotional, guaranteed only through Nov 21, 2026.
OpenAI $4.00 $20.00 ~1.05M tokens Balanced platform.openai.com/docs — Pricing ↗
Optional "fast mode" (research preview) available at $10 / $50, first-party API only.
Anthropic $5.00 $25.00 1M tokens Flagship platform.claude.com/docs — Pricing ↗
Cache hits priced at 0.025x base input (vs 0.1x standard) — the lowest cache rate Anthropic offers.
Anthropic $10.00 $50.00 1M tokens Flagship platform.claude.com/docs — Pricing ↗
Short-context tier shown; long-context requests (very large prompts) bill at $20 / $75.
OpenAI $10.00 $50.00 ~1.05M tokens Flagship platform.openai.com/docs — Pricing ↗

How the prices stack up

Full comparison page →
CHEAPEST INPUT RATE
Mistral · $0.15 in / $0.60 out per million tokens. Good starting point for high-volume, lower-complexity work — classification, extraction, simple summarization.
AVERAGE ACROSS ALL TRACKED MODELS
$2.03 / $9.73
The mean input/output rate across all 25 models we track, flagship and budget tiers included — a rough benchmark for "is this quote reasonable," not a recommendation.
PRICIEST INPUT RATE
Claude Fable 5.1 tied with GPT-6 Astra
Anthropic · $10.00 in / $50.00 out per million tokens — 66.7x the cheapest tracked rate. We don't track benchmark scores, so treat this as a price comparison, not a capability ranking.

How we verify every price

Full methodology →
01
We read the source
Every price on this page comes from the provider's own public pricing page — never a third-party aggregator, a scraped API response, or an estimate. Where a model is open-weight (like Llama), we name the specific host whose price we're showing, since different hosts charge different rates for the same model.
02
We date every check
This is a new tracker — the table above reflects a single verification pass on September 10, 2026. We re-check every price every 48 hours; any change gets a new date and gets logged rather than silently overwritten.
03
We keep it free and honest
No account needed to browse. We're not sponsored by, or affiliated with, any provider listed here — every row is sourced and treated the same way. Prices shown are standard short-context rates; most providers also offer caching, batch, or long-context tiers we link to but don't fully expand in the table.

Frequently asked questions

All FAQs →

Mistral Small 4 (Mistral) is the cheapest model we track, at $0.15 per million input tokens and $0.60 per million output tokens. Several other budget-tier models sit close behind it — see the full table above.

Wide. Claude Fable 5.1 (Anthropic) charges 66.7x more per input token than Mistral Small 4 (tied with GPT-6 Astra at the top end). Flagship pricing generally reflects larger models with deeper reasoning, not a linear improvement in output quality — worth benchmarking against your own task before assuming the expensive model is the right one.

Input tokens are what you send the model — your prompt, context, documents, conversation history. Output tokens are what it generates back. Output is almost always priced higher, often 3–5x the input rate, because generating text is more computationally expensive than reading it.

Meta doesn't sell first-party API access to Llama — it publishes the model weights and lets other companies host it. The prices we show for Llama 4 are from Together AI, one of several inference providers serving it; other hosts (Groq, Fireworks, and others) may price it differently.

Every price on this page was checked directly against the provider's own pricing page on September 10, 2026. We re-verify every 48 hours going forward — this was the first pass, so there's no history yet, but every future check will be dated and any change logged, so the record builds from here.

No. PerTokens is independent and unsponsored. Every provider is listed the same way, using the same source (their own public pricing page) and the same verification date.