Best LLM under $5 per million tokens
- budget, per M output
- $5
- models fit
- 85
- lead over #2
- +0.4
Overall index · read October 22
Cheapest output, $ per million tokens
- S Llama 3 8B Lunaris $0.05
- Gemma 3 4B $0.075
- Mistral Small 3 $0.08
- Llama 3.1 8B Instruct $0.08
- N Nex-N2.5-Mini $0.10
- Reka Edge $0.10
- Ministral 3 3B 2512 $0.10
- Granite 4.0 Micro $0.112
As of October 22, 2026, Google Gemini 3.8 Flash leads 85 models that qualify, with 69.0 on the overall index, ahead of Meta Muse Spark 1.3 (68.6) and Xiaomi MiMo-V2.6-Pro (68.4).
| Rank | Model | Overall index | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|
| 1 | Gemini 3.8 Flash Google | 69.0 | 61.6 | $3.75 | |
| 2 | Muse Spark 1.3 Meta | 68.6 | 57.4 | $4.25 | |
| 3 | MiMo-V2.6-Pro Xiaomi | 68.4 | 65.4 | $0.87 | |
| 4 | Gemini 3.6 Flash Google | 68.3 | 62.5 | $3.75 | |
| 5 | GLM 5.3 Z.ai | 67.9 | 70.6 | $4.40 | |
| 6 | Claude Haiku 5.5 Anthropic | 67.6 | 70.7 | $0.50 | |
| 7 | DeepSeek V4.1 Flash 1017 DeepSeek | 67.6 | 70.3 | $0.60 | |
| 8 | Muse Spark 1.1 Meta | 67.0 | 70.3 | $4.25 | |
| 9 | GLM 5.3 Flash Z.ai | 66.0 | 57.7 | $0.50 | |
| 10 | Gemini 3.7 Flash Google | 65.7 | 58.0 | $3.75 | |
| 11 | GLM 5.2 Z.ai | 62.5 | 61.1 | $4.40 | |
| 12 | DeepSeek V4.1 Flash DeepSeek | 62.4 | 62.5 | $0.84 | |
| 13 | Qwen3.7 Max Qwen | 61.7 | 54.1 | $4.425 | |
| 14 | GPT-5.6 Luna OpenAI | 61.6 | 58.0 | $1.20 | |
| 15 | DeepSeek V4 Pro 0813 DeepSeek | 60.8 | 56.1 | $1.98 | |
| 16 | MiMo-V2.6-Flash Xiaomi | 60.8 | 54.7 | $0.28 | |
| 17 | Step 5 Preview StepFun | 60.5 | 61.1 | $2.70 | |
| 18 | Kimi K2.6 Moonshot AI | 60.5 | 52.1 | $4.00 | |
| 19 | Muse Spark 1.2 Meta | 60.4 | 55.7 | $4.25 | |
| 20 | GPT-6 Luna OpenAI | 59.7 | 57.3 | $0.50 |
How this list is built
Generated from the dataset on every refresh. Nobody picks the order by hand.
Models with a paid output price of $5 or less per million tokens on OpenRouter, ordered by the LLMBenchmark Index, highest first. The index is our own 0 to 100 composite of our five skill indexes, the Artificial Analysis Intelligence Index and the LMArena text Elo, with a 90% margin.
- OpenRouter, fetched
- LMArena WebDev, fetched
- Artificial Analysis (via OpenRouter), fetched