Qwen3.8 27B NVFP4
- Input
- —
- Output
- —
- Speed
- ~150tok/sest.
- Context
- 262Ktokens
Rankings
Coding
Lab-reported#48 / 137rank by quality58.6±10.2quality / 1001/7 benchmarks measuredWriting & chat
Lab-reported#16 / 162rank by quality77.0±9.7quality / 1001/5 benchmarks measuredVision
Lab-reported#28 / 98rank by quality68.1±8.7quality / 1001/6 benchmarks measuredHard reasoning
Lab-reported#51 / 218rank by quality67.2±9.7quality / 1001/4 benchmarks measured
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× B200 180GB
- NVFP4 weights
- Throughput
- ~5,918 tok/s
- many requests batched
- Price
- $0.14–1.22 /M tok
- busy → light use
- On your own machine
- RTX 5090 32GB (NVFP4, ~80 tok/s single-stream)
- single consumer GPU or Mac
Assumptions (5)
- NVFP4 weights (27.8B params) + 25% KV-cache headroom ≈ 20 GB
- 1× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
- ~5,918 output tok/s aggregate at batch 169 (bandwidth-bound)
- Blended 3:1 input:output; prefill ~38,242 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
~#6 of 21 (estimated)~79.4%
- #1Claude Opus 596.0%
- #2Claude Sonnet 585.2%
- #5MiniMax-M380.5%
- ~#6Qwen3.8 27B NVFP4Estimated79.4%
- #6Hy378.0%
No published score — estimated from related benchmarks
SWE-bench Pro (Public)Estimated
~#11 of 24 (estimated)~62.0%
- #1Claude Fable 5.293.0%
- #2Claude Opus 5.589.9%
- #10Claude Sonnet 563.2%
- ~#11Qwen3.8 27B NVFP4Estimated62.0%
- #11Nex-N2.5-Pro61.2%
No published score — estimated from related benchmarks
DeepSWE v1.1Estimated
~#37 of 46 (estimated)~53.0%
- #1Gemini 4 Argon78.8%
- #2DeepSeek V4.1 Flash 101777.0%
- #36Nex-N2.5-Pro55.8%
- ~#37Qwen3.8 27B NVFP4Estimated53.0%
- #37Qwen3.8 Max (0902)51.0%
No published score — estimated from related benchmarks
Terminal-Bench 4.0Estimated
~#29 of 102 (estimated)~23.9%
- #1Kimi K3.1 Pro67.2%
- #2Claude Sonnet 5.563.6%
- #28Qwen3.8 Flash-Next25.3%
- ~#29Qwen3.8 27B NVFP4Estimated23.9%
- #29MiMo-V2.6-Flash22.7%
No published score — estimated from related benchmarks
LiveCodeBenchEstimated
~#11 of 28 (estimated)~81.9%
- #1Qwen3.8-Omni-Flash92.6%
- #2Qwen3.8 Flash-Next91.9%
- #10Ternary Bonsai 27B82.8%
- ~#11Qwen3.8 27B NVFP4Estimated81.9%
- #11Gemma 4 31B80.0%
No published score — estimated from related benchmarks
SciCodeLab-reported
#48 of 11848.4%
- #1Claude Fable 5.267.7%
- #2Claude Opus 5.566.9%
- #47Hy348.6%
- #48Qwen3.8 27B NVFP448.4%
- #49Quasar 438B48.1%
huggingface.co · 2026-09-29
AA Coding Agent IndexEstimated
~#18 of 26 (estimated)~43.5
- #1Claude Opus 5.561.7
- #2Claude Opus 559.7
- #17Muse Spark 1.345.7
- ~#18Qwen3.8 27B NVFP4Estimated43.5
- #18DeepSeek V4 Pro (0813)43.1
No published score — estimated from related benchmarks
Writing & chat
IFBenchLab-reported
#14 of 13978.9%
- #1MiniMax-M382.9%
- #2Qwen3.8 2.4T A95B82.8%
- #13Nova 2.0 Pro Preview79.0%
- #14Qwen3.8 27B NVFP478.9%
- #15Qwen3.5 397B A17B78.8%
huggingface.co · 2026-09-29
LMArena Text, Non-English (Elo)Estimated
~#17 of 57 (estimated)~1456 Elo
- #1Gemini 4 ArgonFew votes1512
- #2Claude Opus 5.51497
- #16DeepSeek V4.1 Flash1459
- ~#17Qwen3.8 27B NVFP4Estimated1456
- #17GPT-5.6 Terra1456
No published score — estimated from related benchmarks
LMArena Text (Elo)Estimated
~#17 of 57 (estimated)~1468 Elo
- #1Gemini 4 Argon1525
- #2Claude Opus 5.51507
- #16DeepSeek V4.1 Flash1474
- ~#17Qwen3.8 27B NVFP4Estimated1468
- #17GPT-5.6 Terra1466
No published score — estimated from related benchmarks
EQ-Bench Creative Writing v3 (Elo)Estimated
~#9 of 26 (estimated)~1974 Elo
- #1GPT-6 Astra2173
- #2Claude Fable 5.12162
- #8Grok 4.72007
- ~#9Qwen3.8 27B NVFP4Estimated1974
- #9Muse Spark 1.31906
No published score — estimated from related benchmarks
Vision
MMMU-ProLab-reported
#39 of 8774.9%
- #1Claude Opus 5.587.7%
- #2GPT-6 Astra86.9%
- #37Qwen3.6 35B A3B75.0%
- #39Qwen3.8 27B NVFP474.9%
- #40Muse Glimmer74.3%
huggingface.co · 2026-09-29
CharXiv ReasoningEstimated
~#6 of 12 (estimated)~81.3%
- #1Gemini 3.8 Flash86.2%
- #2Kimi K384.8%
- #5MiniMax-M3.1-Flash82.5%
- ~#6Qwen3.8 27B NVFP4Estimated81.3%
- #6Nemotron 3.5 Super VL80.8%
No published score — estimated from related benchmarks
OmniDocBench v1.5Estimated
~#4 of 8 (estimated)~87.1%
- #1MiniMax-M391.6%
- #2Kimi K391.1%
- #2Qwen3.8 27B91.1%
- ~#4Qwen3.8 27B NVFP4Estimated87.1%
- #4Gemini 3.1 Pro85.3%
No published score — estimated from related benchmarks
GDP.pdf (AA)Estimated
~#20 of 48 (estimated)~18.5%
- #1Muse Wonder Full32.6%
- #2GPT-6 Astra31.0%
- #19MiMo-V2.6-Pro19.2%
- ~#20Qwen3.8 27B NVFP4Estimated18.5%
- #20Gemini 3.1 Pro17.8%
No published score — estimated from related benchmarks
ChartographyEstimated
~#10 of 25 (estimated)~30.3%
- #1Claude Fable 5.273.4%
- #2GPT-6 Astra71.0%
- #9GPT-5.531.0%
- ~#10Qwen3.8 27B NVFP4Estimated30.3%
- #10GPT-6 Luna29.1%
No published score — estimated from related benchmarks
Hard reasoning
GPQA DiamondLab-reported
#45 of 19288.0%
- #1GPT-6 Astra96.1%
- #2Gemini 3.8 Flash95.3%
- #44DeepSeek-V4-Pro-0813-nvfp4-DSpark88.4%
- #45Qwen3.8 27B NVFP488.0%
- #46Qwen3.8 27B MXFP487.7%
huggingface.co · 2026-09-29
Humanity's Last ExamEstimated
~#57 of 199 (estimated)~30.4%
- #1Claude Fable 5.274.3%
- #2Claude Opus 5.561.4%
- #56Nemotron 3.5 Super VL30.7%
- ~#57Qwen3.8 27B NVFP4Estimated30.4%
- #57A.X-K229.6%
No published score — estimated from related benchmarks
AIME (latest)Estimated
~#2 of 34 (estimated)~98.3%
No published score — estimated from related benchmarks
AA Intelligence IndexEstimated
~#37 of 103 (estimated)~34.2
- #1Gemini 4 Carbon59.4
- #2Claude Opus 5.557.6
- #36DeepSeek V4 Pro (0813)36.0
- ~#37Qwen3.8 27B NVFP4Estimated34.2
- #37Qwen3.8 27B33.7
No published score — estimated from related benchmarks