Intelligence Per Dollar
Benchmark points per blended dollar across every credibly ranked model. Scores are min-max normalized across 176 models with 4+ independent benchmarks; cost assumes a typical 3:1 input:output token mix. Leader = 100.
This is not a quality ranking. #1 here means the most benchmark points per dollar, so a mid-scoring budget model will outrank a frontier model that costs 100x more. Check the Score and % of top score columns for raw capability, and use the task rankings when output quality is what compounds in your workflow.
Value leaderboard
| # | Model | Score | % of top score | In / Out per 1M | Blended $/1M | Value |
|---|---|---|---|---|---|---|
| 1 | OpenAI: gpt-oss-120b BEST VALUEopenai/gpt-oss-120b | 40.2 | 40% | $0.02 / $0.09 | $0.04 | 100.0 |
| 2 | OpenAI: gpt-oss-120b (batch)openai/gpt-oss-120b:batch | 40.2 | 40% | $0.03 / $0.14 | $0.06 | 69.0 |
| 3 | OpenAI: gpt-oss-20b (batch)openai/gpt-oss-20b:batch | 22.5 | 23% | $0.02 / $0.11 | $0.05 | 47.1 |
| 4 | OpenAI: GPT-5 Nano (batch)openai/gpt-5-nano:batch | 32.0 | 32% | $0.03 / $0.20 | $0.07 | 44.9 |
| 5 | Mistral: Mistral Small 3mistralai/mistral-small-24b-instruct-2501 | 26.4 | 27% | $0.05 / $0.08 | $0.06 | 44.3 |
| 6 | Google: Gemma 4 26B A4B google/gemma-4-26b-a4b-it | 48.4 | 49% | $0.07 / $0.23 | $0.11 | 43.7 |
| 7 | Mistral: Mistral Small 3.2 24Bmistralai/mistral-small-3.2-24b-instruct | 45.4 | 46% | $0.07 / $0.20 | $0.11 | 41.2 |
| 8 | OpenAI: gpt-oss-20bopenai/gpt-oss-20b | 22.5 | 23% | $0.03 / $0.13 | $0.06 | 39.4 |
| 9 | OpenAI: GPT-5.6 Luna (batch)openai/gpt-5.6-luna:batch | 89.4 | 90% | $0.10 / $0.60 | $0.23 | 38.3 |
| 10 | OpenAI: GPT-5.4 Nano (batch)openai/gpt-5.4-nano:batch | 68.1 | 69% | $0.10 / $0.62 | $0.23 | 28.4 |
| 11 | StepFun: Step 3.5 Flashstepfun/step-3.5-flash | 42.4 | 43% | $0.10 / $0.30 | $0.15 | 27.2 |
| 12 | NVIDIA: Nemotron 3 Supernvidia/nemotron-3-super-120b-a12b | 39.4 | 40% | $0.07 / $0.40 | $0.16 | 24.3 |
| 13 | Qwen: Qwen3 32Bqwen/qwen3-32b | 32.7 | 33% | $0.08 / $0.28 | $0.13 | 24.2 |
| 14 | Z.ai: GLM 4.7 Flashz-ai/glm-4.7-flash | 34.7 | 35% | $0.06 / $0.40 | $0.15 | 23.0 |
| 15 | OpenAI: GPT-5 Nanoopenai/gpt-5-nano | 32.0 | 32% | $0.05 / $0.40 | $0.14 | 22.4 |
| 16 | Google: Gemma 4 31Bgoogle/gemma-4-31b-it | 53.7 | 54% | $0.20 / $0.40 | $0.25 | 20.7 |
| 17 | MiniMax: MiniMax M3 (batch)minimax/minimax-m3:batch | 78.0 | 79% | $0.22 / $0.88 | $0.39 | 19.4 |
| 18 | OpenAI: GPT-5.6 Lunaopenai/gpt-5.6-luna | 89.4 | 90% | $0.20 / $1.20 | $0.45 | 19.2 |
| 19 | MiniMax: MiniMax M2.7minimax/minimax-m2.7 | 65.2 | 66% | $0.21 / $0.84 | $0.37 | 17.1 |
| 20 | Qwen: Qwen3.6 35B A3Bqwen/qwen3.6-35b-a3b | 52.9 | 53% | $0.10 / $0.90 | $0.30 | 17.0 |
| 21 | Qwen: Qwen3 235B A22B Instruct 2507qwen/qwen3-235b-a22b-2507 | 52.4 | 53% | $0.14 / $0.80 | $0.31 | 16.6 |
| 22 | DeepSeek: DeepSeek V3deepseek/deepseek-chat | 78.9 | 79% | $0.32 / $0.89 | $0.46 | 16.4 |
| 23 | Qwen: Qwen3.5-9Bqwen/qwen3.5-9b | 31.3 | 32% | $0.17 / $0.25 | $0.19 | 15.9 |
| 24 | Qwen: Qwen3.7 Plus (batch)qwen/qwen3.7-plus:batch | 64.8 | 65% | $0.23 / $0.92 | $0.40 | 15.5 |
| 25 | OpenAI: GPT-4.1 Nano (batch)openai/gpt-4.1-nano:batch | 13.7 | 14% | $0.05 / $0.20 | $0.09 | 15.1 |
| 26 | DeepSeek: DeepSeek V4 Pro 0423deepseek/deepseek-v4-pro | 84.4 | 85% | $0.43 / $0.87 | $0.54 | 15.0 |
| 27 | Google: Gemma 3 4Bgoogle/gemma-3-4b-it | 9.7 | 10% | $0.05 / $0.10 | $0.06 | 15.0 |
| 28 | MiniMax: MiniMax M3minimax/minimax-m3 | 78.0 | 79% | $0.30 / $1.20 | $0.52 | 14.3 |
| 29 | OpenAI: GPT-5.4 Nanoopenai/gpt-5.4-nano | 68.1 | 69% | $0.20 / $1.25 | $0.46 | 14.2 |
| 30 | Qwen: Qwen3.5-35B-A3Bqwen/qwen3.5-35b-a3b | 47.9 | 48% | $0.14 / $1.00 | $0.35 | 13.0 |
Budget champions : 80+ score, cheapest first
| # | Model | Score | % of top score | In / Out per 1M | Blended $/1M | Value |
|---|---|---|---|---|---|---|
| 1 | OpenAI: GPT-5.6 Luna (batch)openai/gpt-5.6-luna:batch | 89.4 | 90% | $0.10 / $0.60 | $0.23 | 38.3 |
| 2 | OpenAI: GPT-5.6 Lunaopenai/gpt-5.6-luna | 89.4 | 90% | $0.20 / $1.20 | $0.45 | 19.2 |
| 3 | DeepSeek: DeepSeek V4 Pro 0423deepseek/deepseek-v4-pro | 84.4 | 85% | $0.43 / $0.87 | $0.54 | 15.0 |
| 4 | Z.ai: GLM 5.2 (batch)z-ai/glm-5.2:batch | 87.0 | 88% | $0.37 / $1.50 | $0.65 | 12.9 |
| 5 | Google: Gemini 3.5 Flash (batch)google/gemini-3.5-flash:batch | 84.2 | 85% | $0.75 / $4.50 | $1.69 | 4.8 |
| 6 | Z.ai: GLM 5.2z-ai/glm-5.2 | 87.0 | 88% | $0.01 / $6.80 | $1.70 | 4.9 |
| 7 | Anthropic: Claude Sonnet 5 (batch)anthropic/claude-sonnet-5:batch | 91.4 | 92% | $1.00 / $5.00 | $2.00 | 4.4 |
| 8 | OpenAI: GPT-5.6 Terra (batch)openai/gpt-5.6-terra:batch | 93.0 | 94% | $1.00 / $6.00 | $2.25 | 4.0 |
| 9 | OpenAI: GPT-5.2 (batch)openai/gpt-5.2:batch | 80.1 | 81% | $0.88 / $7.00 | $2.41 | 3.2 |
| 10 | OpenAI: GPT-5.4 (batch)openai/gpt-5.4:batch | 89.6 | 90% | $1.25 / $7.50 | $2.81 | 3.1 |
Free models with credible scores
Per-dollar math breaks at $0. These are simply the strongest free options:
- NVIDIA: Nemotron 3 Ultra (free) : score 64.2
- Google: Gemma 4 31B (free) : score 53.7
- Google: Gemma 4 26B A4B (free) : score 48.4
Assumptions
Value = blended benchmark score divided by blended price per million tokens, indexed to the leader. A 3:1 input:output ratio fits most chat and RAG workloads; estimate your exact mix with the cost calculator. Scoring details in the methodology. Models with fewer than 4 independent benchmarks are excluded rather than guessed.