Live LLM API Price, Value & Promo Index
Real prices, quality-per-dollar value and live deals for LLM APIs — every figure carries its source and the date it was checked. Prices are read at the source, not resold: this is an index, not a gateway.
Knowledge cutoff April 2026. Specs sourced here · full model page →
Knowledge cutoff March 2026. Specs sourced here · full model page →
Latest releases
- Claude Fable 5.1 (Anthropic)2026-09-01$60 /1M
- Granite 4.2 8B (CoreWeave)2026-08-31$0.25 /1M
- GLM 5.3 Flash (Relace)2026-08-26$0.309 /1M
- DeepSeek V4 Flash Vision Exp (DeepInfra)2026-08-21$0.862 /1M
- Qwen3.8 27B (Parasail)2026-08-14$2.44 /1M
- GLM 5.3 (Reka)2026-08-14$4.65 /1M
- Gemini 3.7 Flash (Google)2026-08-13$4.5 /1M
- Grok 4.6 (xAI)2026-08-12$8 /1M
- Qwen3.8 2.4T A95B (Novita)2026-08-12$8 /1M
- Nemotron 3.5 Lightning (DeepInfra)2026-08-11$0.28 /1M
- DeepSeek V4 Flash 0731 (Baidu)-23%$0.195 → $0.15 /1M · 2026-09-05
- DeepSeek V4 (Novita)-16%$5.28 → $4.435 /1M · 2026-09-05
- GLM 5.2 (Baidu)-13%$1.705 → $1.479 /1M · 2026-09-05
- Qwen3 235B A22B Instruct 2507 (Google)-12%$1.25 → $1.1 /1M · 2026-09-05
- Qwen3.8 2.4T A95B (Together)-9%$8.75 → $8 /1M · 2026-09-05
Live free-credit deals
- DeepSeek — $5 free credits on signup$5no carduntil 2026-12-31 · verified 2026-09-02
- Groq — Free developer tier, no credit cardno cardongoing · verified 2026-09-02
Cheapest & best value
Best value (Elo per $) · 52 models
| Model | Elo | /1M | Elo / $ |
|---|---|---|---|
| deepseek-v4-flash | 1436 | $0.15 | 9,573 |
| gemma-4-26b-a4b | 1438 | $0.4 | 3,595 |
| gemma-4-31b | 1451 | $0.44 | 3,298 |
| glm-4.7-flash | 1366 | $0.46 | 2,970 |
| nemotron-3-super-120b-a12b | 1361 | $0.485 | 2,806 |
Cheapest frontier model API · 10 models
| Model | In /1M | Out /1M | Blended |
|---|---|---|---|
| GLM 5.3 Flash (Relace) | $0.0713 | $0.2375 | $0.309 |
| Gemini 3.8 Flash (Google) | $0.75 | $3.75 | $4.5 |
| GLM 5.3 (Reka) | $1.15 | $3.5 | $4.65 |
| GPT-5.6 Sol (OpenAI) | $1 | $5 | $6 |
| GPT-5.6 Terra (OpenAI) | $1 | $6 | $7 |
Cheapest LLM chat API · 108 models
| Model | In /1M | Out /1M | Blended |
|---|---|---|---|
| DeepSeek V4 Flash 0731 (Baidu) | $0.05 | $0.1 | $0.15 |
| Ministral 3 3B 2512 (Mistral) | $0.1 | $0.1 | $0.2 |
| Qwen3.5-9B (SiliconFlow) | $0.1 | $0.15 | $0.25 |
| Nemotron 3 Nano 30B A3B (Crusoe) | $0.05 | $0.2 | $0.25 |
| Granite 4.2 8B (CoreWeave) | $0.1 | $0.15 | $0.25 |
The market, in six charts
Analyses recomputed from the index at every rebuild — not opinion, arithmetic.
The price trend
Every model by release date and price, on one chart.
The falling price of intelligence
What $1 of measured reasoning buys, over time.
When the big model is worth it
The efficient frontier of price × intelligence.
The price of the frontier
Every intelligence record-holder, and its price today.
How fast models get replaced
Release cadence per lab — how long a model stays current.
The age discount
Why last season’s model costs a fraction of today’s.
Cost guides
Practical, data-derived reads on where the bill really goes.