Model performance on MindsHub

See how models served through MindsHub Inference perform: response times, availability and cost, with current prices, API aliases and fallback routes.

Measurements refresh every 2 to 6 hours · About the data

All makers
Makers
Reasoning effort

Performance

Compare the wait before an answer starts, the time to finish it and how fast the model writes. Measured through MindsHub Inference.

Showing 46 models

Measurements: Oct 14 – Nov 17, 2026 UTC.

Complex task · first answer token. Wait for the first answer token on a complex writing task, including thinking time. Each model’s range across effort levels. Ordered by the selected effort. Dots show P50 (typical results).

  • Claude Haiku 4.5no effort settingClaude Haiku 4.5 · no effort setting0.87 sP05 · low1.04 sP50 · typical1.26 sP95 · highMeasurementComplex taskTests40IntervalNov 9–171.04 s
  • GPT-OSS 120B (Fireworks)no effort settingGPT-OSS 120B (Fireworks) · no effort setting1.46 sP05 · low1.89 sP50 · typical4.63 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 171.89 s
  • MindsHub Aireffort: none to mediumMindsHub Air · effort: none1.19 sP05 · low1.43 sP50 · typical2.52 sP95 · highMeasurementComplex taskTests40IntervalOct 20–Nov 17MindsHub Air · effort: low1.69 sP05 · low2.24 sP50 · typical3.76 sP95 · highMeasurementComplex taskTests40IntervalOct 20–Nov 17MindsHub Air · effort: medium2.51 sP05 · low3.42 sP50 · typical4.52 sP95 · highMeasurementComplex taskTests40IntervalOct 20–Nov 171.43–3.42 s, none 1.43 s, low 2.24 s, medium 3.42 s
  • Grok 4.5effort: high onlyGrok 4.5 · effort: high5.44 sP05 · low8.16 sP50 · typical16.8 sP95 · highMeasurementComplex taskTests40IntervalNov 4–178.16 s, low not measured, medium not measured, high 8.16 s
  • MindsHub Blazeeffort: none to highMindsHub Blaze · effort: none0.55 sP05 · low0.77 sP50 · typical1.20 sP95 · highMeasurementComplex taskTests40IntervalOct 29–Nov 17MindsHub Blaze · effort: high5.33 sP05 · low8.59 sP50 · typical14.1 sP95 · highMeasurementComplex taskTests40IntervalNov 9–170.77–8.59 s, none 0.77 s, high 8.59 s
  • Gemini 3.1 Flash-Liteeffort: minimal to highGemini 3.1 Flash-Lite · effort: minimal1.09 sP05 · low1.38 sP50 · typical1.87 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17Gemini 3.1 Flash-Lite · effort: low1.41 sP05 · low1.65 sP50 · typical1.98 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17Gemini 3.1 Flash-Lite · effort: medium4.01 sP05 · low5.04 sP50 · typical9.45 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17Gemini 3.1 Flash-Lite · effort: high7.04 sP05 · low8.71 sP50 · typical12.0 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 171.38–8.71 s, minimal 1.38 s, low 1.65 s, medium 5.04 s, high 8.71 s
  • Muse Ember 1.0no effort settingMuse Ember 1.0 · no effort setting6.70 sP05 · low11.9 sP50 · typical16.7 sP95 · highMeasurementComplex taskTests40IntervalNov 10–1111.9 s
  • Muse Spark 1.1no effort settingMuse Spark 1.1 · no effort setting7.96 sP05 · low15.3 sP50 · typical33.0 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 1715.3 s
  • Gemini 3 Flash Previeweffort: high onlyGemini 3 Flash Preview · effort: high9.43 sP05 · low16.1 sP50 · typical30.8 sP95 · highMeasurementComplex taskTests40IntervalOct 14–Nov 1716.1 s, minimal not measured, low not measured, medium not measured, high 16.1 s
  • GPT-6 Lunaeffort: none to maxGPT-6 Luna · effort: none1.10 sP05 · low1.31 sP50 · typical2.21 sP95 · highMeasurementComplex taskTests40IntervalNov 4–17GPT-6 Luna · effort: low1.65 sP05 · low2.23 sP50 · typical3.55 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17GPT-6 Luna · effort: medium2.35 sP05 · low3.51 sP50 · typical4.40 sP95 · highMeasurementComplex taskTests40IntervalNov 4–17GPT-6 Luna · effort: high3.58 sP05 · low4.68 sP50 · typical7.79 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17GPT-6 Luna · effort: xhigh4.13 sP05 · low6.20 sP50 · typical9.30 sP95 · highMeasurementComplex taskTests40IntervalOct 29–Nov 17GPT-6 Luna · effort: max7.66 sP05 · low16.6 sP50 · typical60.4 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 171.31–16.6 s, none 1.31 s, low 2.23 s, medium 3.51 s, high 4.68 s, xhigh 6.20 s, max 16.6 s
  • Kimi K3no effort settingKimi K3 · no effort setting14.9 sP05 · low21.5 sP50 · typical52.6 sP95 · highMeasurementComplex taskTests40IntervalNov 4–1721.5 s
  • GPT 5.3 Codexeffort: low to xhighGPT 5.3 Codex · effort: low1.48 sP05 · low1.84 sP50 · typical2.63 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17GPT 5.3 Codex · effort: medium2.14 sP05 · low3.58 sP50 · typical6.94 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17GPT 5.3 Codex · effort: high5.84 sP05 · low10.8 sP50 · typical16.6 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17GPT 5.3 Codex · effort: xhigh16.4 sP05 · low21.7 sP50 · typical41.0 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 171.84–21.7 s, low 1.84 s, medium 3.58 s, high 10.8 s, xhigh 21.7 s
  • Muse Spark 1.2no effort settingMuse Spark 1.2 · no effort setting9.76 sP05 · low22.0 sP50 · typical38.5 sP95 · highMeasurementComplex taskTests40IntervalNov 9–1722.0 s
  • DeepSeek V4.1 Flash 1017effort: none to maxDeepSeek V4.1 Flash 1017 · effort: none0.88 sP05 · low1.09 sP50 · typical1.68 sP95 · highMeasurementComplex taskTests40IntervalNov 10–17DeepSeek V4.1 Flash 1017 · effort: low6.26 sP05 · low9.63 sP50 · typical20.4 sP95 · highMeasurementComplex taskTests40IntervalOct 28–Nov 17DeepSeek V4.1 Flash 1017 · effort: high6.50 sP05 · low11.4 sP50 · typical25.3 sP95 · highMeasurementComplex taskTests40IntervalOct 27–Nov 17DeepSeek V4.1 Flash 1017 · effort: max11.1 sP05 · low23.1 sP50 · typical43.2 sP95 · highMeasurementComplex taskTests40IntervalOct 28–Nov 171.09–23.1 s, none 1.09 s, low 9.63 s, high 11.4 s, max 23.1 s
  • GPT 5.4 Nanoeffort: none to xhighGPT 5.4 Nano · effort: none0.85 sP05 · low1.08 sP50 · typical1.76 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17GPT 5.4 Nano · effort: low0.91 sP05 · low1.22 sP50 · typical1.67 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17GPT 5.4 Nano · effort: medium1.27 sP05 · low3.78 sP50 · typical5.14 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17GPT 5.4 Nano · effort: high2.34 sP05 · low4.09 sP50 · typical6.43 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17GPT 5.4 Nano · effort: xhigh11.1 sP05 · low25.0 sP50 · typical41.3 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 171.08–25.0 s, none 1.08 s, low 1.22 s, medium 3.78 s, high 4.09 s, xhigh 25.0 s
  • GLM 5.3 Flasheffort: low to maxGLM 5.3 Flash · effort: low0.91 sP05 · low1.17 sP50 · typical3.13 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17GLM 5.3 Flash · effort: high1.07 sP05 · low1.43 sP50 · typical3.48 sP95 · highMeasurementComplex taskTests40IntervalNov 10–17GLM 5.3 Flash · effort: max8.69 sP05 · low37.7 sP50 · typical82.5 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 171.17–37.7 s, low 1.17 s, high 1.43 s, max 37.7 s
  • Gemini 3.1 Pro Previeweffort: low to highGemini 3.1 Pro Preview · effort: low11.3 sP05 · low28.5 sP50 · typical34.4 sP95 · highMeasurementComplex taskTests40IntervalNov 4–17Gemini 3.1 Pro Preview · effort: medium26.9 sP05 · low33.8 sP50 · typical47.4 sP95 · highMeasurementComplex taskTests40IntervalOct 29–Nov 17Gemini 3.1 Pro Preview · effort: high33.9 sP05 · low44.7 sP50 · typical82.8 sP95 · highMeasurementComplex taskTests40IntervalNov 4–1728.5–44.7 s, low 28.5 s, medium 33.8 s, high 44.7 s
  • Kimi K3.1 Prono effort settingKimi K3.1 Pro · no effort setting32.2 sP05 · low53.4 sP50 · typical74.3 sP95 · highMeasurementComplex taskTests40IntervalOct 27–Nov 1753.4 s
  • Muse Spark 1.3no effort settingMuse Spark 1.3 · no effort setting30.0 sP05 · low55.8 sP50 · typical110 sP95 · highMeasurementComplex taskTests40IntervalNov 9–1755.8 s
  • Gemini 3.8 Flasheffort: low to highGemini 3.8 Flash · effort: low1.97 sP05 · low2.50 sP50 · typical33.4 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17Gemini 3.8 Flash · effort: medium28.6 sP05 · low34.8 sP50 · typical41.2 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17Gemini 3.8 Flash · effort: high40.7 sP05 · low59.7 sP50 · typical100 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 172.50–59.7 s, low 2.50 s, medium 34.8 s, high 59.7 s
  • Muse Wonder 1.0no effort settingMuse Wonder 1.0 · no effort setting33.7 sP05 · low60.9 sP50 · typical103 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 1760.9 s
  • GPT-6.1 Soleffort: low to maxGPT-6.1 Sol · effort: low1.32 sP05 · low1.57 sP50 · typical2.30 sP95 · highMeasurementComplex taskTests40IntervalNov 4–17GPT-6.1 Sol · effort: medium2.29 sP05 · low2.98 sP50 · typical4.89 sP95 · highMeasurementComplex taskTests40IntervalNov 4–17GPT-6.1 Sol · effort: high7.65 sP05 · low10.4 sP50 · typical14.0 sP95 · highMeasurementComplex taskTests40IntervalNov 4–17GPT-6.1 Sol · effort: xhigh18.7 sP05 · low23.6 sP50 · typical34.5 sP95 · highMeasurementComplex taskTests40IntervalNov 4–17GPT-6.1 Sol · effort: max51.3 sP05 · low68.6 sP50 · typical97.7 sP95 · highMeasurementComplex taskTests40IntervalNov 4–171.57–68.6 s, low 1.57 s, medium 2.98 s, high 10.4 s, xhigh 23.6 s, max 68.6 s
  • GLM 5.3effort: low to maxGLM 5.3 · effort: low0.86 sP05 · low1.31 sP50 · typical2.08 sP95 · highMeasurementComplex taskTests40IntervalNov 4–17GLM 5.3 · effort: high1.07 sP05 · low1.47 sP50 · typical2.39 sP95 · highMeasurementComplex taskTests40IntervalNov 4–17GLM 5.3 · effort: max28.1 sP05 · low69.5 sP50 · typical121 sP95 · highMeasurementComplex taskTests40IntervalNov 4–171.31–69.5 s, low 1.31 s, high 1.47 s, max 69.5 s
  • GPT 5.4 Minieffort: none to xhighGPT 5.4 Mini · effort: none0.98 sP05 · low1.12 sP50 · typical1.50 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17GPT 5.4 Mini · effort: low1.14 sP05 · low1.36 sP50 · typical1.65 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17GPT 5.4 Mini · effort: medium1.67 sP05 · low2.34 sP50 · typical3.75 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17GPT 5.4 Mini · effort: high3.80 sP05 · low6.69 sP50 · typical18.0 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17GPT 5.4 Mini · effort: xhigh36.0 sP05 · low69.5 sP50 · typical130 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 171.12–69.5 s, none 1.12 s, low 1.36 s, medium 2.34 s, high 6.69 s, xhigh 69.5 s
  • Qwen3.8-2.4T-A95Beffort: low to xhighQwen3.8-2.4T-A95B · effort: low2.10 sP05 · low2.55 sP50 · typical5.10 sP95 · highMeasurementComplex taskTests40IntervalNov 10–17Qwen3.8-2.4T-A95B · effort: medium3.61 sP05 · low4.29 sP50 · typical16.1 sP95 · highMeasurementComplex taskTests40IntervalNov 10–17Qwen3.8-2.4T-A95B · effort: xhigh37.4 sP05 · low70.5 sP50 · typical138 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 172.55–70.5 s, low 2.55 s, medium 4.29 s, xhigh 70.5 s
  • Gemini 4 Argoneffort: low to highGemini 4 Argon · effort: low2.72 sP05 · low4.76 sP50 · typical35.6 sP95 · highMeasurementComplex taskTests40IntervalOct 17–Nov 17Gemini 4 Argon · effort: medium2.71 sP05 · low4.24 sP50 · typical28.2 sP95 · highMeasurementComplex taskTests40IntervalOct 17–Nov 17Gemini 4 Argon · effort: high60.9 sP05 · low79.6 sP50 · typical125 sP95 · highMeasurementComplex taskTests40IntervalOct 17–Nov 174.24–79.6 s, low 4.76 s, medium 4.24 s, high 79.6 s
  • GPT-6 Astraeffort: low to maxGPT-6 Astra · effort: low1.59 sP05 · low2.01 sP50 · typical3.16 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17GPT-6 Astra · effort: medium3.86 sP05 · low5.78 sP50 · typical8.59 sP95 · highMeasurementComplex taskTests40IntervalNov 10–17GPT-6 Astra · effort: high11.4 sP05 · low15.3 sP50 · typical20.3 sP95 · highMeasurementComplex taskTests40IntervalNov 4–17GPT-6 Astra · effort: xhigh30.5 sP05 · low41.0 sP50 · typical51.8 sP95 · highMeasurementComplex taskTests40IntervalNov 4–17GPT-6 Astra · effort: max68.9 sP05 · low90.5 sP50 · typical128 sP95 · highMeasurementComplex taskTests40IntervalNov 9–172.01–90.5 s, low 2.01 s, medium 5.78 s, high 15.3 s, xhigh 41.0 s, max 90.5 s
  • Claude Fable 5.1effort: low to maxClaude Fable 5.1 · effort: low1.08 sP05 · low1.62 sP50 · typical3.45 sP95 · highMeasurementComplex taskTests40IntervalOct 14–Nov 17Claude Fable 5.1 · effort: medium1.42 sP05 · low2.14 sP50 · typical4.08 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 17Claude Fable 5.1 · effort: high1.68 sP05 · low2.79 sP50 · typical13.0 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 17Claude Fable 5.1 · effort: xhigh13.0 sP05 · low15.8 sP50 · typical46.8 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 17Claude Fable 5.1 · effort: max62.1 sP05 · low92.5 sP50 · typical156 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 171.62–92.5 s, low 1.62 s, medium 2.14 s, high 2.79 s, xhigh 15.8 s, max 92.5 s
  • Grok 4.7effort: low to xhighGrok 4.7 · effort: low3.00 sP05 · low5.44 sP50 · typical9.47 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17Grok 4.7 · effort: medium5.70 sP05 · low9.77 sP50 · typical37.7 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17Grok 4.7 · effort: high9.25 sP05 · low38.7 sP50 · typical250 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 17Grok 4.7 · effort: xhigh10.8 sP05 · low135 sP50 · typical405 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 175.44–135 s, low 5.44 s, medium 9.77 s, high 38.7 s, xhigh 135 s
  • Claude Haiku 5.5effort: low to maxClaude Haiku 5.5 · effort: low1.91 sP05 · low2.36 sP50 · typical3.02 sP95 · highMeasurementComplex taskTests40IntervalNov 13–14Claude Haiku 5.5 · effort: medium2.33 sP05 · low2.89 sP50 · typical5.30 sP95 · highMeasurementComplex taskTests40IntervalNov 6–17Claude Haiku 5.5 · effort: high3.97 sP05 · low5.85 sP50 · typical10.7 sP95 · highMeasurementComplex taskTests40IntervalNov 13–14Claude Haiku 5.5 · effort: xhigh9.61 sP05 · low13.6 sP50 · typical17.5 sP95 · highMeasurementComplex taskTests40IntervalNov 13–14Claude Haiku 5.5 · effort: max101 sP05 · low136 sP50 · typical199 sP95 · highMeasurementComplex taskTests40IntervalNov 13–142.36–136 s, low 2.36 s, medium 2.89 s, high 5.85 s, xhigh 13.6 s, max 136 s
  • GPT-6.1 Terraeffort: none to maxGPT-6.1 Terra · effort: none1.31 sP05 · low1.53 sP50 · typical2.27 sP95 · highMeasurementComplex taskTests40IntervalOct 29–Nov 17GPT-6.1 Terra · effort: low1.48 sP05 · low1.88 sP50 · typical2.61 sP95 · highMeasurementComplex taskTests40IntervalOct 29–Nov 17GPT-6.1 Terra · effort: medium2.19 sP05 · low3.00 sP50 · typical5.32 sP95 · highMeasurementComplex taskTests40IntervalOct 29–Nov 17GPT-6.1 Terra · effort: high5.20 sP05 · low7.31 sP50 · typical9.79 sP95 · highMeasurementComplex taskTests40IntervalOct 29–Nov 17GPT-6.1 Terra · effort: xhigh22.5 sP05 · low28.9 sP50 · typical47.3 sP95 · highMeasurementComplex taskTests40IntervalOct 29–Nov 17GPT-6.1 Terra · effort: max113 sP05 · low155 sP50 · typical213 sP95 · highMeasurementComplex taskTests40IntervalOct 29–Nov 171.53–155 s, none 1.53 s, low 1.88 s, medium 3.00 s, high 7.31 s, xhigh 28.9 s, max 155 s
  • Claude Sonnet 5.5effort: low to maxClaude Sonnet 5.5 · effort: low1.23 sP05 · low1.60 sP50 · typical2.32 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 17Claude Sonnet 5.5 · effort: medium1.26 sP05 · low1.65 sP50 · typical2.59 sP95 · highMeasurementComplex taskTests40IntervalOct 29–Nov 17Claude Sonnet 5.5 · effort: high3.35 sP05 · low4.45 sP50 · typical7.58 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 17Claude Sonnet 5.5 · effort: xhigh5.43 sP05 · low7.78 sP50 · typical12.2 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 17Claude Sonnet 5.5 · effort: max108 sP05 · low174 sP50 · typical249 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 171.60–174 s, low 1.60 s, medium 1.65 s, high 4.45 s, xhigh 7.78 s, max 174 s
  • Claude Opus 5.5effort: low to maxClaude Opus 5.5 · effort: low1.39 sP05 · low1.74 sP50 · typical2.26 sP95 · highMeasurementComplex taskTests40IntervalNov 10–17Claude Opus 5.5 · effort: medium2.78 sP05 · low3.68 sP50 · typical5.01 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 17Claude Opus 5.5 · effort: high3.88 sP05 · low6.89 sP50 · typical11.9 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 17Claude Opus 5.5 · effort: xhigh27.8 sP05 · low40.9 sP50 · typical56.5 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 17Claude Opus 5.5 · effort: max240 sP05 · low313 sP50 · typical403 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 171.74–313 s, low 1.74 s, medium 3.68 s, high 6.89 s, xhigh 40.9 s, max 313 s
  • Claude Sonnet 5effort: high onlyClaude Sonnet 5 · effort: high1.20 sP05 · low1.34 sP50 · typical1.64 sP95 · highMeasurementComplex taskTests40IntervalNov 9–171.34 s, low not measured, medium not measured, high 1.34 s, max not measured
  • GPT 5.6 Terraeffort: medium onlyGPT 5.6 Terra · effort: medium1.49 sP05 · low1.76 sP50 · typical2.17 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 171.76 s, none not measured, low not measured, medium 1.76 s, high not measured, xhigh not measured, max not measured
  • GPT 5.6 Lunaeffort: medium onlyGPT 5.6 Luna · effort: medium1.58 sP05 · low1.91 sP50 · typical2.51 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 171.91 s, none not measured, low not measured, medium 1.91 s, high not measured, xhigh not measured, max not measured
  • GPT-6 Soleffort: medium onlyGPT-6 Sol · effort: medium1.76 sP05 · low2.30 sP50 · typical3.17 sP95 · highMeasurementComplex taskTests40IntervalNov 4–172.30 s, none not measured, low not measured, medium 2.30 s, high not measured, xhigh not measured, max not measured
  • GPT 5.6 Soleffort: medium onlyGPT 5.6 Sol · effort: medium2.25 sP05 · low2.97 sP50 · typical4.28 sP95 · highMeasurementComplex taskTests40IntervalNov 4–172.97 s, none not measured, low not measured, medium 2.97 s, high not measured, xhigh not measured, max not measured
  • Claude Fable 5effort: high onlyClaude Fable 5 · effort: high2.68 sP05 · low3.13 sP50 · typical3.86 sP95 · highMeasurementComplex taskTests40IntervalOct 16–Nov 173.13 s, low not measured, medium not measured, high 3.13 s, xhigh not measured, max not measured
  • Claude Opus 5effort: high onlyClaude Opus 5 · effort: high1.90 sP05 · low3.28 sP50 · typical5.56 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 173.28 s, low not measured, medium not measured, high 3.28 s, xhigh not measured, max not measured
  • Claude Fable 5.2effort: high onlyClaude Fable 5.2 · effort: high2.09 sP05 · low5.37 sP50 · typical8.66 sP95 · highMeasurementComplex taskTests40IntervalNov 10–115.37 s, low not measured, medium not measured, high 5.37 s, xhigh not measured, max not measured
  • Gemini 3.6 Flasheffort: medium onlyGemini 3.6 Flash · effort: medium20.4 sP05 · low26.4 sP50 · typical34.2 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 1726.4 s, minimal not measured, low not measured, medium 26.4 s, high not measured
  • Gemini 3.5 Flasheffort: medium onlyGemini 3.5 Flash · effort: medium20.3 sP05 · low28.2 sP50 · typical37.6 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 1728.2 s, minimal not measured, low not measured, medium 28.2 s, high not measured
  • DeepSeek V4.1 Flasheffort: high onlyDeepSeek V4.1 Flash · effort: high21.7 sP05 · low32.7 sP50 · typical54.6 sP95 · highMeasurementComplex taskTests40IntervalNov 10–1732.7 s, none not measured, low not measured, high 32.7 s, max not measured
  • Gemini 3.7 Flasheffort: medium onlyGemini 3.7 Flash · effort: medium29.2 sP05 · low34.4 sP50 · typical44.0 sP95 · highMeasurementComplex taskTests40IntervalOct 15–Nov 1734.4 s, low not measured, medium 34.4 s, high not measured
  • Grok 4.6effort: high onlyGrok 4.6 · effort: high18.6 sP05 · low76.0 sP50 · typical141 sP95 · highMeasurementComplex taskTests40IntervalNov 4–1776.0 s, low not measured, medium not measured, high 76.0 s, xhigh not measured

Bar: from the fastest of a model’s effort levels to the slowest. Dot: one effort level, its typical test run. Ring: the toolbar’s effort level. Current models are tested at each reasoning effort; older versions at their default effort.

About the data

What does this page show?

Performance and reliability measured through MindsHub Inference, alongside current token prices, API aliases and fallback routes. The figures describe the models as served through MindsHub. They are not a ranking of every provider’s deployment of those models.

Where do the measurements come from?

We run automated API checks and writing tasks through MindsHub Inference. Complex task, writing speed and cost per task figures refresh every two hours, and the other measurements every six hours. Complex task performance and cost cover up to 35 days of tests. Each chart shows its measurement dates. If an update fails, we keep previously saved measurements with their original dates. Prices, aliases and fallback routes come from the same live catalog as our pricing page. A dash means no usable measurement is available.

How is Ping measured?

Every 10 minutes, checks from multiple regions ask for a one-word reply at the lowest reasoning effort. We time the first answer token, including any thinking before it. Ping measures basic responsiveness and tracks availability.

Sample prompt

Reply with the single word: ok
How is the Simple task measured?

Every six hours, the model receives a short briefing about a fictional market hall and is asked to describe a Saturday morning in at least 150 words. It must interpret the facts, decide how to present them and follow the writing constraints. We time the first answer token at the default reasoning effort, including thinking time.

Sample prompt, abridged; briefing omitted

Using only the supplied market-hall briefing, describe a typical Saturday morning in at least 150 words. Use plain paragraphs, with no headings or lists, and do not add facts.
How is the Complex task measured?

The model receives a longer briefing about the same fictional market hall and must turn it into a complete visitor’s guide of at least 800 words. It must analyze and organize the supplied facts, cover them in order and obey content and formatting constraints. We measure the first answer token, total response time and writing speed at each tested effort level. The first-answer and total times include thinking; writing speed starts when the answer begins. Repeated tests run over several weeks; each model’s details show the test count and interval. These are speed tests, not scores for reasoning quality.

Sample prompt, abridged; briefing omitted

Using only the supplied briefing, write a visitor’s guide of at least 800 words. Cover every part in the given order, in plain paragraphs with no headings, lists or tables. Do not add facts.
What are p50, p05 and p95?

P50 is the median: half of measurements are at or below it and half at or above it. P05 and P95 mark the lower and upper ends of the middle 90% of measurements. The leaderboard shows p50. Open Performance to explore the spread; Ping and Simple task have p50 and p95 only.

Does availability below 100% mean requests failed?

Not necessarily. A configured fallback route can still answer when the model you chose does not. Availability before fallback is the share of our automated checks answered by that model, including retries, over the selected period. Other checks were answered by a fallback or got no answer. These checks do not measure how often customer requests fail. Failures attributed to our monitoring network are excluded.

What does the intelligence score mean?

MindsHub’s read is a subjective estimate from 1 to 10 based on public benchmarks and rankings, not a precise measurement. Performance depends on your workload; test models on your own tasks.

For detailed benchmarks, refer to third party services such as Artificial Analysis, Epoch AI or similar.

How is cost per task compared?

The most expensive measured model and effort level is the 100% reference, updated automatically as results change. The leaderboard range covers a model’s measured effort levels; expand it for individual values. In Cost per task, choose another reference to compare against. Dollar estimates use mean input and output tokens at current standard rates, excluding cache discounts. Your bill depends on your requests.

What changes with reasoning effort?

Intelligence, Complex task timings and writing speed follow the selected effort. Ping always uses the lowest effort and Simple task the default. The cost range covers all measured levels. Views showing all effort levels disable the selector because every level is already visible.

What are fixed and rotating aliases?

An alias is the name sent in the API’s model field. Fixed aliases select a specific model version. Rotating aliases follow a model family as it updates. MindsHub Air and MindsHub Blaze are managed rotating aliases: MindsHub can change the model behind each name. The Aliases tab explains their purpose and lists the names you can use.

How do token prices, context limits and the free tier work?

Tokens are pieces of text. The context limit is how much input a model accepts in one request. Token prices are US dollars per million input or output tokens, from the same catalog as the pricing page. MindsHub Air is free within the free-tier limits; its listed token rates apply to usage beyond those limits. TypeSafe’s Jev is free. See Prices for cached-input, cache-write and other rates.

Which models are included?

The leaderboard covers chat models and their managed aliases. Other models lists decision and embedding models with their own response-time measurements and availability. See mindshub.ai/unified-inference#free-tier for the free tier and mindshub.ai/pricing for rates and credits.