Model Pricing
USD per 1M tokens, grouped by weight class so open-weight heavies stand next to the big labs at their real price points. Thinking tokens bill as output. $/task = mean published cost-per-completed-task (τ-bench, Terminal-Bench, reasoning suites) — never our own estimate. Every number carries its source. Updated 2026-07-12 · vendor pages are the source of truth · pricepertoken cross-checked.
Frontier The judgment tier — best reasoning money buys
| Model | In /1M ↕ | Out /1M ↕ | Cached | $/task | Context | GPQA ↕ | Capabilities | Src |
|---|---|---|---|---|---|---|---|---|
| Grok 4.5xAI ⚠ EU access pending AI Act compliance (xAI: mid-July 2026) |
$2.00 | $6.00 | $0.50 | — | 500K | — | textcodingtool-callingagentic | ↗ |
| Gemini 3.1 ProGoogle | $2.00 | $12.00 | $0.20 | — | 1000K | 94.1 | textcodingtool-callingagenticmultimodalvoice | ↗ |
| Claude Opus 4.8Anthropic | $5.00 | $25.00 | $0.50 | — | 1000K | 92.0 | textcodingtool-callingagenticmultimodal | ↗ |
| GPT-5.6 SolOpenAI | $5.00 | $30.00 | — | — | 400K | — | textcodingtool-callingagenticmultimodal | ↗ |
| Claude Fable 5Anthropic | $10.00 | $50.00 | $1.00 | — | 1000K | 92.6 | textcodingtool-callingagenticmultimodal | ↗ |
Heavy Near-frontier capability, workhorse pricing
| Model | In /1M ↕ | Out /1M ↕ | Cached | $/task | Context | GPQA ↕ | Capabilities | Src |
|---|---|---|---|---|---|---|---|---|
| Kimi K2.6Moonshot AI ⚠ listed free (promo) at seed time — verify |
free* | free* | $0.14 | — | 256K | 78.8 | textcodingtool-callingagenticopen-weights | ↗ |
| MiniMax M2.7MiniMax ⚠ listed free (promo) at seed time — verify |
free* | free* | $0.054 | — | 205K | 87.4 | textcodingtool-callingopen-weights | ↗ |
| Gemini 3 FlashGoogle | $0.50 | $3.00 | $0.05 | — | 1000K | 89.8 | textcodingtool-callingmultimodalvoice | ↗ |
| GLM-5.2Z.AI ⚠ RXed eval winner 12/07/2026 — closest to Fable on audit work at ~9x lower cost |
$1.40 | $4.40 | $0.26 | — | 1000K | — | textcodingtool-callingagenticopen-weights | ↗ |
| Gemini 3.5 FlashGoogle | $1.50 | $9.00 | $0.15 | — | 1000K | 82.8 | textcodingtool-callingmultimodal | ↗ |
| Claude Sonnet 4.6Anthropic | $3.00 | $15.00 | $0.30 | — | 1000K | 79.7 | textcodingtool-callingagenticmultimodal | ↗ |
| GPT-5.6 TerraOpenAI | $2.50 | $15.00 | — | — | 400K | — | textcodingtool-callingagenticmultimodal | ↗ |
Medium Production volume work
| Model | In /1M ↕ | Out /1M ↕ | Cached | $/task | Context | GPQA ↕ | Capabilities | Src |
|---|---|---|---|---|---|---|---|---|
| Grok 4.1 FastxAI ⚠ free at listing time — verify before relying |
free* | free* | — | — | 2000K | 85.3 | textcodingtool-calling | ↗ |
| DeepSeek Chat (V4)DeepSeek ⚠ aggregator price unusually low — verify on vendor page |
$0.014 | $0.028 | — | — | 164K | — | textcodingtool-callingopen-weights | ↗ |
| GPT-OSS 120BOpenAI (open) | $0.036 | $0.10 | $0.04 | — | 131K | 67.2 | textcodingtool-callingopen-weights | ↗ |
| DeepSeek V4 FlashDeepSeek | $0.09 | $0.18 | $0.003 | — | 1000K | 71.6 | textcodingtool-callingopen-weights | ↗ |
| Grok 4 FastxAI | $0.20 | $0.50 | — | — | 2000K | 84.7 | textcodingtool-calling | ↗ |
| Llama 4 MaverickMeta (hosted) | $0.15 | $0.60 | $0.17 | — | 1000K | 67.1 | textcodingmultimodalopen-weights | ↗ |
| Qwen3 VL 235BAlibaba Qwen | $0.20 | $0.88 | $0.10 | — | 262K | 71.2 | textcodingmultimodalopen-weights | ↗ |
| Gemini 3.1 Flash LiteGoogle | $0.25 | $1.50 | $0.025 | — | 1000K | 82.2 | textcodingmultimodal | ↗ |
| Grok 4.3xAI | $1.25 | $2.50 | $0.20 | — | — | 65.8 | textcodingtool-calling | ↗ |
| Claude Haiku 4.5Anthropic | $1.00 | $5.00 | $0.10 | — | 200K | 67.2 | textcodingtool-callingmultimodal | ↗ |
| GPT-5.6 LunaOpenAI | $1.00 | $6.00 | — | — | 400K | — | textcodingtool-callingmultimodal | ↗ |
Light High-volume, simple tasks
*free = listed at $0 by the aggregator at scrape time (usually launch promos) — verify before relying. No affiliate links. Weight classes assigned from published capability-index scores (GPQA / Artificial Analysis), not marketing tiers.
What would YOUR workload cost? tokens in/out per day → $/month, every model
| Model | Class | $/day | $/month |
|---|
Straight per-token math on the listed prices — no caching, batching or thinking-token overhead. Treat as a floor for reasoning models (thinking tokens bill as output).