All models
Live speed benchmarks for every model across all providers. Numbers update continuously — sorted by latest tokens per second within each provider. New to comparing model speed? Read the Ollama speed comparison explainer.
opencode-go-anthropic
| Model | TPS now | TPS 24h avg | TTFT | Reliability | Intelligence Index |
|---|---|---|---|---|---|
| 1 Claude Haiku 5 5 | 632.8 | 901.1 | 2.0s | 75% | 43.4 |
| 2 GPT 6 Luna | 351.0 | — | 3.7s | 0% | 38.1 |
| 3 Step 5 Preview | 209.4 | 156.2 | 891ms | 100% | 43.7 |
| 4 MiniMax M3.1 Flash | 184.2 | 106.1 | 751ms | 100% | 42.7 |
| 5 Beam (Free) | 181.3 | 167.5 | 624ms | 100% | — |
| 6 DeepSeek V4.1 Flash | 141.2 | 130.3 | 1.1s | 100% | 39.5 |
| 7 MiniMax M3 | 132.4 | 87.1 | 578ms | 100% | 29.2 |
| 8 Hy3 | 131.2 | 110.0 | 4.6s | 100% | 25.3 |
| 9 GPT 5.6 Luna | 118.5 | 183.8 | 1.5s | 50% | 37.3 |
| 10 GLM 5.3 | 111.3 | 97.7 | 1.6s | 100% | 44.8 |
| 11 GLM 5.2 | 103.9 | 100.1 | 1.4s | 100% | 33.7 |
| 12 Iceman Contributor (Free) | 100.7 | — | 4.2s | 0% | — |
| 13 DeepSeek V4 Pro | 95.9 | 61.3 | 1.3s | 100% | 36 |
| 14 Grok 4.7 | 71.0 | 73.1 | 2.9s | 100% | 46.4 |
| 15 GLM 5.3 Flash | 66.1 | 54.2 | 1.1s | 100% | 41.8 |
| 16 Qwen3.7 Plus | 62.5 | 62.5 | 1.4s | 50% | 25.2 |
| 17 Longcat 2.0 | 58.6 | 57.2 | 1.9s | 100% | 19.1 |
| 18 Qwen3.8 Flash | 56.0 | 59.3 | 1.2s | 100% | 39.8 |
| 19 Kimi K2.7 Code | 46.8 | 42.5 | 1.2s | 100% | 25.8 |
| 20 MiMo V2.6 Pro | 41.2 | 57.1 | 2.6s | 100% | 46.3 |
| 21 Qwen3.8 Max | 41.2 | 42.5 | 1.9s | 100% | 45.4 |
| 22 Kimi K3 | 34.4 | 46.4 | 16.5s | 100% | 43.6 |
| 23 MiMo V2.6 Flash | 32.3 | 25.5 | 1.7s | 100% | 37.9 |
| 24 Hy4 Preview | 28.5 | 24.2 | 1.9s | 100% | — |
| 25 Kimi K3.1 Pro | 28.4 | 29.8 | 2.7s | 100% | — |
ollama-free
| Model | TPS now | TPS 24h avg | TTFT | Reliability | Intelligence Index |
|---|---|---|---|---|---|
| 1 GPT-OSS 120B | 349.4 | 316.7 | 676ms | 100% | 11.6 |
| 2 Nemotron 3 Nano 30B (non-reasoning) | 291.6 | 214.4 | 303ms | 96% | 6.8 |
| 3 Gemma4 31B | 271.8 | 225.4 | 344ms | 100% | 14.7 |
| 4 Nemotron 3 Super | 214.5 | 176.1 | 2.8s | 100% | 12.8 |
| 5 GPT-OSS 20B | 133.4 | 145.4 | 437ms | 100% | 9 |
| 6 Beam | 100.1 | 161.2 | 458ms | 100% | 30.3 |
| 7 Nemotron 3 Ultra | 31.8 | 44.7 | 632ms | 100% | 22.9 |
ollama
| Model | TPS now | TPS 24h avg | TTFT | Reliability | Intelligence Index |
|---|---|---|---|---|---|
| 1 DeepSeek V4.1 Flash | 322.9 | 234.6 | 405ms | 100% | 39.5 |
| 2 Muse Spark Open | 322.5 | 204.0 | 1.9s | 99% | 38.9 |
| 3 Nemotron 3 Nano 30B (non-reasoning) | 317.5 | 194.6 | 364ms | 100% | 6.8 |
| 4 MiMo V2.6 Pro | 314.8 | 244.6 | 400ms | 100% | 46.3 |
| 5 Gemma4 31B | 274.7 | 237.4 | 449ms | 100% | 14.7 |
| 6 MiMo V2.6 Flash | 256.7 | 249.1 | 321ms | 100% | 37.9 |
| 7 DeepSeek V4.1 Flash 1017 | 225.9 | 230.2 | 404ms | 100% | 45.1 |
| 8 GPT-OSS 120B | 175.0 | 276.6 | 357ms | 99% | 11.6 |
| 9 GPT-OSS 20B | 167.4 | 145.8 | 409ms | 99% | 9 |
| 10 DeepSeek V4 Pro 0813 | 162.1 | 142.0 | 474ms | 100% | 36 |
| 11 Nemotron 3 Super | 161.5 | 160.7 | 471ms | 100% | 12.8 |
| 12 GLM 5.2 | 160.1 | 175.4 | 518ms | 100% | 33.7 |
| 13 Kimi K2.6 | 144.8 | 82.0 | 631ms | 100% | 27 |
| 14 GLM 5.3 | 134.9 | 118.1 | 709ms | 100% | 44.8 |
| 15 Mistral Large 4 | 133.1 | 108.6 | 882ms | 100% | 38.4 |
| 16 GLM 5.3 Flash | 123.6 | 150.9 | 692ms | 100% | 41.8 |
| 17 Kimi K2.7 Code | 118.5 | 121.4 | 673ms | 100% | 25.8 |
| 18 MiniMax M3 | 115.8 | 97.5 | 701ms | 99% | 29.2 |
| 19 Beam | 104.5 | 168.5 | 470ms | 99% | 30.3 |
| 20 Mistral Large 3 675B (non-reasoning) | 88.4 | 80.5 | 679ms | 100% | 9.3 |
| 21 Kimi K3 | 86.8 | 105.4 | 1.3s | 100% | 43.6 |
| 22 Nemotron 3 Ultra | 70.6 | 45.7 | 576ms | 100% | 22.9 |
| 23 MiniMax M2.7 | 54.6 | 42.7 | 690ms | 100% | 22.8 |
| 24 MiniMax M3.1 Flash | 37.6 | 118.0 | 728ms | 100% | 42.7 |
opencode-go-responses
| Model | TPS now | TPS 24h avg | TTFT | Reliability | Intelligence Index |
|---|---|---|---|---|---|
| 1 Muse Spark 1.2 Contributor | 226.0 | 211.7 | 1.5s | 100% | 39.6 |
| 2 Muse Spark 1.3 Contributor | 104.3 | 198.4 | 2.2s | 100% | 48.1 |
opencode-zen
| Model | TPS now | TPS 24h avg | TTFT | Reliability | Intelligence Index |
|---|---|---|---|---|---|
| 1 Nemotron 3.5 Lightning (Free) | 57.0 | — | 651ms | 0% | 12.9 |
| 2 Big Pickle | 50.3 | — | 697ms | 0% | — |
opencode-zen-responses
| Model | TPS now | TPS 24h avg | TTFT | Reliability | Intelligence Index |
|---|---|---|---|---|---|
| 1 Muse Spark 1.3 Contributor (Free) | 54.4 | — | 6.3s | 0% | 48.1 |
10 models not currently benchmarked
- GLM 5.3 Flash (Free) — opencode-zen
- Pulse Alpha (Free) — opencode-zen
- DeepSeek V4.1 Flash (Free) — opencode-zen
- Nemotron 3.5 Super Vl (Free) — opencode-zen
- Beam (Free) — opencode-zen
- Gemini 4 Argon — opencode-go-anthropic
- Flint 40B (Free) — opencode-zen
- MiMo V2.6 Flash (Free) — opencode-zen
- Muse Spark 1.2 Contributor — opencode-go-anthropic
- Muse Spark 1.3 Contributor — opencode-go-anthropic
How these numbers are measured
Every model is benchmarked with the same prompt and the same measurement method, regardless of provider. That makes the numbers directly comparable. Sampling cadence does vary by provider — about every 10 minutes on Ollama Pro, about every 60 minutes on the capped plans — but cadence affects only how fresh a number is, not how it is measured. Use the compare tool to overlay speed timelines for up to 6 models, including the same model on two different providers. See the methodology page for the full measurement spec.