G9v3-39A5B
- Input
- —
- Output
- —
- Speed
- ~100tok/sest.
- Context
- 131Ktokens
Long-tail import from Artificial Analysis (2026-09-24); AA variant: G9v3-39A5B. Price = AA median host price.
Rankings
Self-hostingEstimated
How we estimate →What it would cost to run these open weights yourself on rented GPUs. No API sells this model, so this is its price on the chart.
- Hardware
- 1× H200 141GB
- FP8 weights
- Throughput
- ~6,662 tok/s
- many requests batched
- Price
- $0.061–0.50 /M tok
- busy → light use
- On your own machine
- RTX 5090 32GB (int4, 300+ tok/s single-stream)
- single consumer GPU or Mac
Assumptions (5)
- FP8 weights (39B params, 5B active per token) + 25% KV-cache headroom ≈ 49 GB
- 1× H200 141GB at $3.59–$7.91/GPU-hour on-demand (2026-09-24)
- ~6,662 output tok/s aggregate at batch 239 (bandwidth-bound); MoE compute scales with active params
- Blended 3:1 input:output; prefill ~93,508 tok/s
- Low = 75% utilization at the low GPU price; high = 20% at the high price
Benchmark scores
~ italic, dashed = no published score yet, estimated from related benchmarks.
Coding
SWE-bench VerifiedEstimated
~#17 of 21 (estimated)~66.9%
- #1Claude Opus 596.0%
- #2Claude Sonnet 585.2%
- #16Nemotron 3 Ultra70.7%
- ~#17G9v3-39A5BEstimated66.9%
- #17Intern-S2-Preview (35B-A3B)64.0%
No published score — estimated from related benchmarks
SWE-bench Pro (Public)Estimated
~#23 of 24 (estimated)~44.4%
- #1Claude Fable 5.293.0%
- #2Claude Opus 5.589.9%
- #22Muse Glimmer51.2%
- ~#23G9v3-39A5BEstimated44.4%
- #23Nex-N2.5-Mini43.8%
No published score — estimated from related benchmarks
DeepSWE v1.1Estimated
~#43 of 46 (estimated)~34.8%
- #1Gemini 4 Argon78.8%
- #2DeepSeek V4.1 Flash 101777.0%
- #42Nex-N2.5-Mini36.1%
- ~#43G9v3-39A5BEstimated34.8%
- #43Hy328.0%
No published score — estimated from related benchmarks
Terminal-Bench 4.0Estimated
~#42 of 102 (estimated)~5.5%
- #1Kimi K3.1 Pro67.2%
- #2Claude Sonnet 5.563.6%
- #41Qwen3.8 27B5.6%
- ~#42G9v3-39A5BEstimated5.5%
- #42Gemini 3.1 Pro4.0%
No published score — estimated from related benchmarks
LiveCodeBenchEstimated
~#17 of 28 (estimated)~61.1%
- #1Qwen3.8-Omni-Flash92.6%
- #2Qwen3.8 Flash-Next91.9%
- #16Claude Haiku 4.561.5%
- ~#17G9v3-39A5BEstimated61.1%
- #17GLM-5.3 NVFP453.5%
No published score — estimated from related benchmarks
SciCodeVerified
#91 of 11836.8%
- #1Claude Fable 5.267.7%
- #2Claude Opus 5.566.9%
- #90Granite 4.2 30B37.8%
- #91G9v3-39A5B36.8%
- #92Mistral Large 336.6%
artificialanalysis.ai · 2026-09-24
AA Coding Agent IndexEstimated
~#25 of 26 (estimated)~27.2
- #1Claude Opus 5.561.7
- #2Claude Opus 559.7
- #24GPT-6 Luna31.1
- ~#25G9v3-39A5BEstimated27.2
- #25Muse Spark Open26.3
No published score — estimated from related benchmarks
Hard reasoning
GPQA DiamondVerified
#78 of 19280.5%
- #1GPT-6 Astra96.1%
- #2Gemini 3.8 Flash95.3%
- #77Qwen3.5 9B80.6%
- #78G9v3-39A5B80.5%
- #79Nemotron 3 Super 120B A12B80.0%
artificialanalysis.ai · 2026-09-24
Humanity's Last ExamVerified
#88 of 19817.5%
- #1Claude Fable 5.274.3%
- #2Claude Opus 5.561.4%
- #87K2 Horizon 7B18.2%
- #88G9v3-39A5B17.5%
- #89HyperNova 60B 260516.6%
artificialanalysis.ai · 2026-09-24
AIME (latest)Estimated
~#21 of 34 (estimated)~81.6%
- #1GLM-5.3 NVFP499.5%
- #2Beam 1.0 501B A23B97.7%
- #20Claude Haiku 4.583.7%
- ~#21G9v3-39A5BEstimated81.6%
- #21Magistral Small 1.280.3%
No published score — estimated from related benchmarks
AA Intelligence IndexEstimated
~#64 of 103 (estimated)~19.9
- #1Gemini 4 Carbon59.4
- #2Claude Opus 5.557.6
- #63Ling 3.0 Flash20.1
- ~#64G9v3-39A5BEstimated19.9
- #64LongCat 2.019.1
No published score — estimated from related benchmarks
Agents
τ²-bench (Telecom)Estimated
~#43 of 120 (estimated)~68.8%
- #1Step 3.7 Flash98.5%
- #2DeepSeek-V4-Pro-0813-nvfp4-DSpark98.3%
- #42Apriel-v1.6-15B-Thinker69.3%
- ~#43G9v3-39A5BEstimated68.8%
- #43Nemotron 3 Super 120B A12B67.8%
No published score — estimated from related benchmarks
OSWorld-VerifiedEstimated
~#11 of 15 (estimated)~67.3%
- #1Kimi K384.8%
- #2Qwen3.8 27B84.3%
- #10Nex-N2.5-Mini71.2%
- ~#11G9v3-39A5BEstimated67.3%
- #11Muse Glimmer65.9%
No published score — estimated from related benchmarks
OSWorld 2.0Estimated
~last of 5 (estimated)below measured range (<47.5%)
No published score — estimated from related benchmarks
BrowseCompEstimated
~#19 of 22 (estimated)~68.0%
- #1Atria Dawn Preview92.5%
- #2GPT-6 Astra91.5%
- #18Nemotron 3.5 Bolt 74B-A9B70.0%
- ~#19G9v3-39A5BEstimated68.0%
- #19Mistral Medium 3.548.6%
No published score — estimated from related benchmarks
GDPval-AA (Elo)Estimated
~#31 of 92 (estimated)~1335 Elo
- #1Claude Opus 5.51866
- #2Muse Wonder Full1857
- #30GPT-5.51336
- ~#31G9v3-39A5BEstimated1335
- #31MiniMax-M31242
No published score — estimated from related benchmarks
AutomationBench-AAEstimated
~#39 of 85 (estimated)~21.3%
- #1Claude Fable 5.276.4%
- #2Muse Wonder Full71.3%
- #38Quasar 438B23.9%
- ~#39G9v3-39A5BEstimated21.3%
- #39MiniMax-M321.3%
No published score — estimated from related benchmarks