Every served model, measured on a schedule through FastMetal's own gateway.
This is a report of what we observe, not an SLA.
| Model | Time to first response (p50) | Throughput (p50) | Prompts not retained |
|---|---|---|---|
| Beam 1.0 501B A23Bbeam-1.0 | 923msn=268 | — | — |
| Claude Fable 5anthropic-claude-fable-5 | 3.6sn=648 | 140 tokens/sn=27 | — |
| Claude Fable 5.1anthropic-claude-fable-5-1 | 3.1sn=657 | 126 tokens/sn=27 | — |
| Claude Haiku 4.5anthropic-claude-haiku-4-5 | 797msn=657 | 357 tokens/sn=27 | — |
| Claude Haiku 5.5anthropic-claude-haiku-5-5 | 1.9sn=409 | 513 tokens/sn=18 | — |
| Claude Opus 4.6anthropic-claude-opus-4-6 | 1.6sn=646 | 146 tokens/sn=27 | — |
| Claude Opus 4.8anthropic-claude-opus-4-8 | 1.5sn=660 | 124 tokens/sn=27 | — |
| Claude Opus 5anthropic-claude-opus-5 | 1.9sn=660 | 733 tokens/sn=27 | — |
| Claude Opus 5.5anthropic-claude-opus-5-5 | 2.2sn=659 | 166 tokens/sn=27 | — |
| Claude Sonnet 4.6anthropic-claude-sonnet-4-6 | 928msn=652 | 378 tokens/sn=26 | — |
| Claude Sonnet 5anthropic-claude-sonnet-5 | 1.8sn=660 | 204 tokens/sn=27 | — |
| Claude Sonnet 5.5anthropic-claude-sonnet-5-5 | 1.6sn=620 | 318 tokens/sn=25 | — |
| DeepSeek V4 Flashdeepseek-v4-flash | 1.1sn=651 | 40 tokens/sn=27 | — |
| DeepSeek V4 Prodeepseek-v4-pro | 707msn=656 | 133 tokens/sn=27 | — |
| DeepSeek V4.1 Flashdeepseek-v4.1-flash | 664msn=660 | 194 tokens/sn=27 | — |
| DeepSeek V4.1 Flash 1017deepseek-v4.1-flash-1017 | 699msn=200 | — | — |
| deepseek-v4-flash-0731deepseek-v4-flash-0731 | 1.1sn=658 | 55 tokens/sn=25 | — |
| Devstral 2 2512devstral-2512 | 1.1sn=652 | 250 tokens/sn=27 | Yes |
| Ember-1ember-1 | 671msn=644 | 85 tokens/sn=17 | Yes |
| Fugu Maxfugu-max | 3.9sn=641 | 1307 tokens/sn=25 | — |
| Fugu Ultra v2fugu-ultra-v2 | 4.4sn=655 | 22222 tokens/sn=24 | — |
| Gemini 3.1 Pro Previewgemini-3.1-pro-preview | 5.2sn=626 | 710 tokens/sn=25 | — |
| Gemini 3.5 Flashgemini-3.5-flash | 1.4sn=660 | 598 tokens/sn=27 | — |
| Gemini 3.7 Flashgemini-3.7-flash | 1.9sn=651 | 1233 tokens/sn=27 | — |
| Gemini 3.8 Flashgemini-3.8-flash | 2.7sn=626 | 1059 tokens/sn=25 | — |
| Gemini 4 Argongemini-4-argon | 3.3sn=293 | 89 tokens/sn=15 | — |
| Gemma 4 26B A4B gemma-4-26b-a4b-it | 378msn=659 | 66 tokens/sn=26 | — |
| Gemma 4 31Bjapan-gemma-4-31b | 1.0sn=659 | 77 tokens/sn=26 | — |
| GLM 4.6Vglm-4.6v | 921msn=659 | — | Yes |
| GLM 4.7glm-4.7 | 909msn=623 | — | — |
| GLM 4.7 Flashglm-4.7-flash | 660msn=658 | — | — |
| GLM 5glm-5 | 2.2sn=658 | — | — |
| GLM 5.2glm-5.2 | 1.6sn=659 | — | — |
| GLM 5.3glm-5.3 | 1.1sn=656 | 85 tokens/sn=15 | — |
| GLM 5.3 Flashglm-5.3-flash | 1.0sn=658 | — | — |
| GLM 5.3 FlashXglm-5.3-flashx | 1.2sn=660 | 337 tokens/sn=20 | Yes |
| GPT-4.1 Minigpt-4.1-mini | 761msn=660 | 154 tokens/sn=27 | — |
| GPT-4.1 Nanogpt-4.1-nano | 749msn=660 | 220 tokens/sn=27 | — |
| GPT-4o-minigpt-4o-mini | 1.3sn=660 | 124 tokens/sn=27 | — |
| GPT-5 Minigpt-5-mini | 1.4sn=660 | — | — |
| GPT-5.4 Nanogpt-5.4-nano | 1.6sn=659 | 119 tokens/sn=27 | — |
| GPT-5.6 Lunagpt-5.6-luna | 1.4sn=660 | 167 tokens/sn=27 | — |
| GPT-5.6 Solgpt-5.6-sol | 1.1sn=658 | 205 tokens/sn=27 | — |
| GPT-5.6 Terragpt-5.6-terra | 1.4sn=660 | 206 tokens/sn=27 | — |
| GPT-6 Astragpt-6-astra | 3.0sn=641 | 111 tokens/sn=24 | — |
| GPT-6 Astra Progpt-6-astra-pro | 3.3sn=658 | 10070 tokens/sn=27 | — |
| GPT-6 Lunagpt-6-luna | 2.6sn=628 | 111 tokens/sn=23 | — |
| GPT-6 Luna Progpt-6-luna-pro | 2.3sn=659 | 9574 tokens/sn=27 | — |
| GPT-6 Solgpt-6-sol | 2.7sn=660 | 99 tokens/sn=24 | — |
| GPT-6 Sol Progpt-6-sol-pro | 2.3sn=660 | 9210 tokens/sn=26 | — |
| GPT-6.1 Solgpt-6.1-sol | 2.0sn=618 | 94 tokens/sn=25 | — |
| GPT-6.1 Sol Progpt-6.1-sol-pro | 3.3sn=618 | 9036 tokens/sn=24 | — |
| GPT-6.1 Terragpt-6.1-terra | — | — | — |
| GPT-6.1 Terra Progpt-6.1-terra-pro | — | — | — |
| gpt-oss-120bgpt-oss-120b | 1.1sn=660 | 325 tokens/sn=26 | — |
| gpt-oss-20bgpt-oss-20b | 410msn=660 | 115 tokens/sn=20 | — |
| Grok 4.5grok-4.5 | 872msn=658 | 134 tokens/sn=27 | — |
| Grok 4.6grok-4.6 | 1.4sn=659 | 97 tokens/sn=27 | — |
| Grok 4.7grok-4.7 | 1.1sn=645 | 202 tokens/sn=26 | — |
| Hy3hy3 | 1.7sn=660 | — | — |
| Inklinginkling | 438msn=648 | 133 tokens/sn=22 | Yes |
| Kimi K2.6kimi-k2.6 | 831msn=636 | 34 tokens/sn=18 | — |
| Kimi K2.7 Codejapan-kimi-k2.7-code | 470msn=660 | 114 tokens/sn=26 | — |
| Kimi K3kimi-k3 | 1.5sn=655 | — | — |
| LFM2.5-24B-A2Blfm-2.5-24b-a2b | 376msn=269 | — | Yes |
| Ling 3.0 Flash VLling-3.0-flash-vl | 1.3sn=599 | 69 tokens/sn=20 | — |
| Llama 3.1 8B Instructllama-3.1-8b-instruct | 2.3sn=657 | 40 tokens/sn=27 | — |
| Llama 3.2 3B Instructllama-3.2-3b-instruct | 3.3sn=658 | 100000 tokens/sn=27 | Yes |
| Llama 3.3 70B Instructllama-3.3-70b-instruct | 2.9sn=651 | 20 tokens/sn=27 | — |
| Llama 4 Maverickllama-4-maverick | 1.9sn=655 | 19 tokens/sn=27 | Yes |
| LLM-jp-3.1 8x13B instruct4llm-jp-3.1-8x13b-instruct4 | 551msn=660 | 124 tokens/sn=27 | — |
| Mercury 2.5mercury-2.5 | 602msn=646 | — | Yes |
| Mercury 3mercury-3 | — | — | Yes |
| MiMo-V2.6-Flashmimo-v2.6-flash | 955msn=640 | 84 tokens/sn=27 | — |
| MiMo-V2.6-Promimo-v2.6-pro | 912msn=657 | 101 tokens/sn=27 | — |
| MiniMax M2.7minimax-m2.7 | 1.2sn=659 | — | — |
| MiniMax M3minimax-m3 | 633msn=603 | 22 tokens/sn=25 | — |
| Mistral Large 4mistral-large-4 | 960msn=441 | — | — |
| Muse Glimmer 30Bmuse-glimmer-30b | 606msn=655 | — | — |
| Muse Spark 1.2muse-spark-1.2 | 956msn=658 | — | — |
| Muse Spark 1.3muse-spark-1.3 | 1.1sn=648 | — | — |
| Muse Spark 1.3 Contributormuse-spark-1.3-contributor | 1.1sn=650 | — | — |
| Muse Spark Openmuse-spark-open | 466msn=265 | — | Yes |
| Muse Wondermuse-wonder | 1.3sn=307 | — | — |
| Nano Banana 2 (Gemini 3.1 Flash Image)google-nano-banana-2 | — | — | — |
| Nemotron 3 Supernemotron-3-super-120b-a12b | 618msn=496 | 78 tokens/sn=23 | — |
| Nemotron 3.5 Super VLnemotron-3.5-super-vl | 451msn=266 | — | — |
| Paretopareto | 1.4sn=657 | 25000 tokens/sn=17 | Yes |
| Qwen3 Next 80B A3B Instructqwen3-next-80b-a3b-instruct | 479msn=655 | 196 tokens/sn=27 | — |
| Qwen3 VL 235B A22B Instructqwen3-vl-235b-a22b-instruct | 799msn=659 | 13 tokens/sn=27 | — |
| Qwen3.6 35B-A3Bjapan-qwen3.6-35b | 313msn=660 | — | — |
| qwen3.6-27bqwen3.6-27b | 498msn=651 | — | — |
| Qwen3.7 Maxqwen3.7-max | 2.4sn=658 | 95 tokens/sn=20 | — |
| Qwen3.8 2.4T A95Bqwen3.8-2.4t-a95b | 1.5sn=659 | 52 tokens/sn=26 | Yes |
| Qwen3.8 27Bqwen3.8-27b | 384msn=635 | 45 tokens/sn=26 | — |
| Qwen3.8 Maxqwen3.8-max | 1.1sn=657 | 70 tokens/sn=27 | — |
| Seedream 4.5bytedance-seedream-4.5 | — | — | Yes |
| Solar Mini 4solar-mini4 | 481msn=658 | 95 tokens/sn=27 | — |
| Solar Pro 4solar-pro4 | 469msn=629 | 82 tokens/sn=25 | — |
| Voxtral Mini 3Bmistral-voxtral-mini-3b-2507 | 416msn=659 | 547 tokens/sn=22 | — |
Last updated: 2026-10-28 10:06 UTC
A dash means we do not yet have enough samples — not that the model is slow. Models appear as their sample count passes the threshold. Uptime is not shown: our sampling rate cannot support a meaningful figure.