LLM API pricing · updated twice daily
Every model price.
Every change.
On the record.
TokensWatch tracks list prices across the major LLM APIs, diffs them against the community dataset every morning, and publishes the delta with a source attached. Right now the cheapest blended price is GPT-5 nano.
Current API pricing
| Model | Provider | Input / MTok | Output / MTok | Blended* | Context | Class |
|---|---|---|---|---|---|---|
| GPT-5 nano | OpenAI | $0.05 | $0.40 | $0.14 | 400K | fast |
| Gemini 2.5 Flash-Lite | $0.10 | $0.40 | $0.18 | 1M | fast | |
| DeepSeek V3.2 | DeepSeek | $0.28 | $0.42 | $0.32 | 128K | workhorse |
| GPT-5 mini | OpenAI | $0.25 | $2 | $0.69 | 400K | workhorse |
| Gemini 2.5 Flash | $0.30 | $2.50 | $0.85 | 1M | workhorse | |
| Grok 4 | xAI | $1.25 | $2.50 | $1.56 | 256K | frontier |
| Grok 4 Fast | xAI | $1.25 | $2.50 | $1.56 | 2M | fast |
| Claude Haiku 4.5 | Anthropic | $1 | $5 | $2 | 200K | fast |
| Mistral Medium 3.5 | Mistral | $1.50 | $7.50 | $3 | 128K | workhorse |
| GPT-5.1 | OpenAI | $1.25 | $10 | $3.44 | 400K | frontier |
| Gemini 3 Pro | $2 | $12 | $4.50 | 1M | frontier | |
| Claude Sonnet 4.5 | Anthropic | $3 | $15 | $6 | 200K | workhorse |
| Claude Opus 4.5 | Anthropic | $5 | $25 | $10 | 200K | frontier |
*Blended assumes a 3:1 input-to-output token ratio, typical for chat and RAG workloads. USD per 1M tokens, first-party API, standard tier — no caching or batch discounts.
Every price is read back against the provider's own page, then diffed against the community pricing dataset twice daily (06:17 & 18:17 UTC).
Tech news
Ignatian Examen on artificial intelligence, Oct. 7
Norfolk Police Department E911 Division implements Artificial Intelligence (AI) system on non-emerge
East Lansing to deploy artificial intelligence for parking enforcement
NYC Council holding rare hearing on AI risks with top tech executives
Latest price movements
-
Grok 4 pricing changed: input $3 → $1.25, output $15 → $2.5 per MTok.
-
Grok 4 Fast pricing changed: input $0.2 → $1.25, output $0.5 → $2.5 per MTok.
-
Mistral replaced Medium 3 with Mistral Medium 3.5 at $1.50/$7.50 per MTok — a 3.75× price increase on the Medium line (Medium 3 was $0.40/$2.00). Verified against mistral.ai/pricing/api.
-
Claude Opus 4.5 launched at $5/$25 per MTok — a 67% cut versus Opus 4.1's $15/$75.
-
DeepSeek V3.2 cut API prices by 50%+ to $0.28/$0.42 per MTok (cache miss).