TOKENAIR DECISION LAB

Compare LLM API pricing and model economics.

Compare official AI API prices per 1M tokens, then examine transparent cost metrics and independently measured coding-workflow evidence.

Compare modelsModel my workflow
SOURCE-BOUND CATALOGCoverage you can audit

Price evidence and workflow benchmarks stay separate until a model has a verifiable match.

53priced models
19benchmark-matched
11providers
Catalog 2026.07.27Review evidence

Blended API price per 1M tokens

Estimated cost for 1M tokens under one disclosed input, cached-input, and output mix. Lower is better; official component prices remain visible in the details.

Blended unit price
Data & methods

Official API prices are stored in USD. CNY values are reference conversions at 1 USD = ¥6.7722 using the ECB rate verified Jul 24, 2026. The blended price uses the disclosed input, cache, and output mix.

Estimated blended price / 1M tokens (USD)

Key findings

  1. Ministral 3 3B from Mistral AI is the lowest estimated blended price in this view at $0.0595 per 1M tokens.
  2. The complete view covers 53 officially priced models from 11 providers.
  3. The blended value is derived from the disclosed mix; input, cached-input, and output prices remain first-party fields.
Lowest blended API pricesOfficial input, cached-input, and output prices · blended price uses the disclosed mix
RankModelProviderEstimated blended price / 1M tokens (USD)Input price / 1M tokensCached input price / 1M tokensOutput price / 1M tokensStatusVerified
1Ministral 3 3BMistral AI$0.0595$0.1$0.01$0.1currentJul 27, 2026
2DeepSeek V4 FlashDeepSeek$0.0923$0.14$0.0028$0.28currentJul 27, 2026
3Hunyuan HY3Tencent Hunyuan$0.1271$0.132$0.033$0.528currentJul 27, 2026
4Hunyuan MT2 PlusTencent Hunyuan$0.13$0.1$0.1$0.4currentJul 27, 2026
5Mistral Small 4Mistral AI$0.1343$0.15$0.015$0.6currentJul 27, 2026
6GPT-5.4 nanoOpenAI$0.224$0.2$0.02$1.25currentJul 27, 2026
7Codestral 25.08Mistral AI$0.2385$0.3$0.03$0.9currentJul 27, 2026
Review formula and sources
Complete API pricing dataShow every model currently included in this chart as crawlable HTML.53
Complete API pricing data
RankModelProviderEstimated blended price / 1M tokens (USD)Input price / 1M tokensCached input price / 1M tokensOutput price / 1M tokensStatusVerifiedOfficial price source
1Ministral 3 3BMistral AI$0.0595$0.1$0.01$0.1currentJul 27, 2026Mistral API pricing
2DeepSeek V4 FlashDeepSeek$0.0923$0.14$0.0028$0.28currentJul 27, 2026DeepSeek API pricing and model limits
3Hunyuan HY3Tencent Hunyuan$0.1271$0.132$0.033$0.528currentJul 27, 2026TokenHub model pricing
4Hunyuan MT2 PlusTencent Hunyuan$0.13$0.1$0.1$0.4currentJul 27, 2026TokenHub model pricing
5Mistral Small 4Mistral AI$0.1343$0.15$0.015$0.6currentJul 27, 2026Mistral API pricing
6GPT-5.4 nanoOpenAI$0.224$0.2$0.02$1.25currentJul 27, 2026GPT-5.4 nano model specification
7Codestral 25.08Mistral AI$0.2385$0.3$0.03$0.9currentJul 27, 2026Mistral API pricing
8MiniMax M2.5MiniMax$0.2685$0.3$0.03$1.2currentJul 27, 2026MiniMax pay-as-you-go pricing
9MiniMax M3MiniMax$0.282$0.3$0.06$1.2currentJul 27, 2026MiniMax pay-as-you-go pricing
10MiniMax M2.7MiniMax$0.282$0.3$0.06$1.2currentJul 27, 2026MiniMax pay-as-you-go pricing
11DeepSeek V4 ProDeepSeek$0.2844$0.435$0.0036$0.87currentJul 27, 2026DeepSeek API pricing and model limits
12Qwen3.6 FlashAlibaba Qwen$0.375$0.25$0.25$1.5currentJul 27, 2026Model Studio model inference pricing
13Mistral Large 3Mistral AI$0.3975$0.5$0.05$1.5currentJul 27, 2026Mistral API pricing
14Gemini 3.5 Flash-LiteGoogle Gemini$0.3985$0.3$0.03$2.5currentJul 27, 2026Gemini Developer API pricing
15Qwen3.7 PlusAlibaba Qwen$0.52$0.4$0.4$1.6currentJul 27, 2026Model Studio model inference pricing
16MiniMax M2.5 HighSpeedMiniMax$0.5235$0.6$0.03$2.4currentJul 27, 2026MiniMax pay-as-you-go pricing
17MiniMax M2.7 HighSpeedMiniMax$0.537$0.6$0.06$2.4currentJul 27, 2026MiniMax pay-as-you-go pricing
18GLM-4.7Zhipu GLM$0.5395$0.6$0.11$2.2currentJul 27, 2026Z.AI API pricing
19Gemini 3 Flash PreviewGoogle Gemini$0.5475$0.5$0.05$3previewJul 27, 2026Gemini Developer API pricing
20Qwen3.5 PlusAlibaba Qwen$0.6$0.4$0.4$2.4currentJul 27, 2026Model Studio model inference pricing
21Kimi K2.5Moonshot Kimi$0.615$0.6$0.1$3currentJul 27, 2026TokenHub model pricing
22Grok Build 0.1xAI$0.74$1$0.2$2currentJul 27, 2026xAI API pricing
23Qwen3.6 PlusAlibaba Qwen$0.75$0.5$0.5$3currentJul 27, 2026Model Studio model inference pricing
24GPT-5.4 miniOpenAI$0.8213$0.75$0.075$4.5currentJul 27, 2026GPT-5.4 mini model specification
25GLM-5Zhipu GLM$0.86$1$0.2$3.2currentJul 27, 2026Z.AI API pricing
26Kimi K2.6Moonshot Kimi$0.8995$0.95$0.16$4currentJul 27, 2026Kimi K2.6 pricing and specification
27Grok 4.3xAI$0.9025$1.25$0.2$2.5currentJul 27, 2026xAI API pricing
28Grok 4.20 Multi-AgentxAI$0.9025$1.25$0.2$2.5currentJul 27, 2026xAI API pricing
29Grok 4.20 ReasoningxAI$0.9025$1.25$0.2$2.5currentJul 27, 2026xAI API pricing
30Grok 4.20 Non-ReasoningxAI$0.9025$1.25$0.2$2.5currentJul 27, 2026xAI API pricing
31Kimi K2.7 CodeMoonshot Kimi$0.913$0.95$0.19$4currentJul 27, 2026Kimi K2.7 Code pricing and specification
32Claude Haiku 4.5Anthropic$0.995$1$0.1$5currentJul 27, 2026Claude API pricing
33GLM-5 TurboZhipu GLM$1.048$1.2$0.24$4currentJul 27, 2026Z.AI API pricing
34GPT-5.6 LunaOpenAI$1.095$1$0.1$6currentJul 27, 2026OpenAI model comparison and API pricing
35GLM-5.2Zhipu GLM$1.187$1.4$0.26$4.4currentJul 27, 2026Z.AI API pricing
36GLM-5.1Zhipu GLM$1.187$1.4$0.26$4.4currentJul 27, 2026Z.AI API pricing
37Gemini 3.6 FlashGoogle Gemini$1.4925$1.5$0.15$7.5currentJul 27, 2026Gemini Developer API pricing
38Mistral Medium 3.5Mistral AI$1.4925$1.5$0.15$7.5currentJul 27, 2026Mistral API pricing
39Grok 4.5xAI$1.635$2$0.3$6currentJul 27, 2026xAI API pricing
40Gemini 3.5 FlashGoogle Gemini$1.6425$1.5$0.15$9currentJul 27, 2026Gemini Developer API pricing
41Kimi K2.7 Code HighSpeedMoonshot Kimi$1.826$1.9$0.38$8currentJul 27, 2026Kimi K2.7 Code pricing and specification
42Claude Sonnet 5Anthropic$1.99$2$0.2$10currentJul 27, 2026Claude API pricing
43Gemini 3.1 Pro PreviewGoogle Gemini$2.19$2$0.2$12previewJul 27, 2026Gemini Developer API pricing
44GPT-5.6 TerraOpenAI$2.7375$2.5$0.25$15currentJul 27, 2026OpenAI model comparison and API pricing
45GPT-5.4OpenAI$2.7375$2.5$0.25$15currentJul 27, 2026GPT-5.4 model specification
46Kimi K3Moonshot Kimi$2.985$3$0.3$15currentJul 27, 2026Kimi K3 pricing and specification
47Qwen3.7 MaxAlibaba Qwen$3$2.5$2.5$7.5currentJul 27, 2026Model Studio model inference pricing
48Claude Opus 5Anthropic$4.975$5$0.5$25currentJul 27, 2026Claude API pricing
49Claude Opus 4.8Anthropic$4.975$5$0.5$25currentJul 27, 2026Claude API pricing
50Claude Opus 4.7Anthropic$4.975$5$0.5$25currentJul 27, 2026Claude API pricing
51GPT-5.6 SolOpenAI$5.475$5$0.5$30currentJul 27, 2026OpenAI model comparison and API pricing
52GPT-5.5OpenAI$5.475$5$0.5$30currentJul 27, 2026GPT-5.5 model specification
53Claude Fable 5Anthropic$9.95$10$1$50currentJul 27, 2026Claude API pricing

Monthly cost vs accepted coding workflows

Project the independently measured Terminal-Bench v2.1 cost and pass@1 rate over one fixed monthly attempt volume. Upper-left is better; thin lines connect comparable models from the same provider.

Data & methods
19/53 models · 10/11 providers
Why this chart has fewer modelsOnly models matched to independent Terminal-Bench evidence are plotted. Price-only models remain in the API unit-price chart until a verifiable benchmark match exists.Coverage method
Provider linesLine = provider · point = model · hover for model name
OpenAI3
Anthropic5
Google Gemini3
xAI1
DeepSeek2
Moonshot Kimi1
Zhipu GLM1
Alibaba Qwen1
MiniMax1
Mistral AI1

Projected accepted Terminal-Bench workflows / 1,000 attempts

Modeled monthly cost for 1,000 benchmark attempts ($, log)

Key findings

  1. DeepSeek V4 Flash has the lowest projected monthly benchmark cost in this view at $9.19 for 1,000 attempts.
  2. Claude Opus 5 has the highest projected accepted-workflow count in this view at 891.
  3. This view includes 19 models from 10 providers with independently matched Terminal-Bench evidence.
Complete workflow benchmark dataShow every benchmark-matched model currently included in this chart as crawlable HTML.19
Complete workflow benchmark data
ModelProviderMonthly costAcceptedPass@1Cost / acceptedBenchmark variantVerifiedIndependent benchmark
DeepSeek V4 FlashDeepSeek$9.1961861.8%$0.0149DeepSeek V4 Flash (max)Jul 27, 2026Terminal-Bench v2.1 benchmark leaderboard
DeepSeek V4 ProDeepSeek$15.0264064%$0.0235DeepSeek V4 Pro (max)Jul 27, 2026Terminal-Bench v2.1 benchmark leaderboard
Gemini 3.5 Flash-LiteGoogle Gemini$42.8553653.6%$0.08Gemini 3.5 Flash-LiteJul 27, 2026Terminal-Bench v2.1 benchmark leaderboard
GPT-5.6 LunaOpenAI$44.1180980.9%$0.0545GPT-5.6 Luna (max)Jul 27, 2026Terminal-Bench v2.1 benchmark leaderboard
MiniMax M3MiniMax$46.965265.2%$0.072MiniMax-M3Jul 27, 2026Terminal-Bench v2.1 benchmark leaderboard
Grok 4.5xAI$82.3681781.6%$0.1009Grok 4.5 (high)Jul 27, 2026Terminal-Bench v2.1 benchmark leaderboard
GLM-5.2Zhipu GLM$95.3677977.9%$0.1224GLM-5.2 (max)Jul 27, 2026Terminal-Bench v2.1 benchmark leaderboard
Gemini 3.1 Pro PreviewGoogle Gemini$96.3373873.8%$0.1306Gemini 3.1 Pro PreviewJul 27, 2026Terminal-Bench v2.1 benchmark leaderboard
GPT-5.6 SolOpenAI$125.4688088%$0.1425GPT-5.6 Sol (max)Jul 27, 2026Terminal-Bench v2.1 benchmark leaderboard
Kimi K3Moonshot Kimi$135.6385085%$0.1595Kimi K3Jul 27, 2026Terminal-Bench v2.1 benchmark leaderboard
Claude Haiku 4.5Anthropic$139.2244244.2%$0.315Claude 4.5 HaikuJul 27, 2026Terminal-Bench v2.1 benchmark leaderboard
GPT-5.6 TerraOpenAI$173.8888088%$0.1976GPT-5.6 Terra (max)Jul 27, 2026Terminal-Bench v2.1 benchmark leaderboard
Gemini 3.6 FlashGoogle Gemini$233.4377577.5%$0.3011Gemini 3.6 FlashJul 27, 2026Terminal-Bench v2.1 benchmark leaderboard
Claude Sonnet 5Anthropic$314.6580580.5%$0.3908Claude Sonnet 5 (max)Jul 27, 2026Terminal-Bench v2.1 benchmark leaderboard
Mistral Medium 3.5Mistral AI$321.3250650.6%$0.6355Mistral Medium 3.5Jul 27, 2026Terminal-Bench v2.1 benchmark leaderboard
Qwen3.7 MaxAlibaba Qwen$380.8974574.5%$0.511Qwen3.7 MaxJul 27, 2026Terminal-Bench v2.1 benchmark leaderboard
Claude Opus 5Anthropic$403.5589189.1%$0.4527Claude Opus 5 (max)Jul 27, 2026Terminal-Bench v2.1 benchmark leaderboard
Claude Fable 5Anthropic$560.0784684.6%$0.6617Claude Fable 5 (with fallback)Jul 27, 2026Terminal-Bench v2.1 benchmark leaderboard
Claude Opus 4.8Anthropic$570.5784684.6%$0.6741Claude Opus 4.8 (max)Jul 27, 2026Terminal-Bench v2.1 benchmark leaderboard
DeepSeek V4 FlashDeepSeek V4 Flash (max)
Monthly cost
$9.19
Accepted
618
Pass@1
61.8%
Cost / accepted
$0.015
Independent benchmark
3/3 models selectedClick a chart point to compare. Selecting a fourth model replaces the oldest.
DeepSeek V4 FlashDeepSeek$9.19 · 618 accepted61.8% pass@1 · $0.015 / Accepted
GPT-5.6 LunaOpenAI$44.11 · 809 accepted80.9% pass@1 · $0.055 / Accepted
GPT-5.6 SolOpenAI$125.46 · 880 accepted88% pass@1 · $0.143 / Accepted
Cost calculator
Scenario calculatorOfficial prices+ your workload assumptions

Calculate my workflow

Estimate cost from real list prices while keeping retries, failures, cache and acceptance visible.

Go to Cost & Routing
Documented compatibility
Cursor
OpenCode
Cline
Aider

Check an integration

Separate documented setup support from browser test results and unverified capabilities.

Go to Compatibility
Source & methodologyPrice fields come from linked official provider pages. The blended unit price is derived from one disclosed input, cache, and output mix. Workflow efficiency uses independent Terminal-Bench v2.1 pass@1 and cost-per-task evidence, projected over the selected attempt volume. It is not a provider quota or a universal production acceptance rate.Learn more