Qwen3.8 27B
- Input
- $0.50/M tok
- Output
- $3.00/M tok
- Blended
- $1.125/M tok · 3:1
- Speed
- 54tok/s
- Context
- 262Ktokens
- Cost per task
- $3.5/task · AA
Open weights.
Rankings
Coding
Mixed#62 / 137rank by quality48.5±5.3quality / 1004/7 benchmarks measuredWriting & chat
Mixed#49 / 162rank by quality61.9±3.6quality / 1004/5 benchmarks measuredVision
Mixed#29 / 98rank by quality67.9±5.4quality / 1003/6 benchmarks measuredHard reasoning
Verified#53 / 218rank by quality65.7±3.7quality / 1003/4 benchmarks measuredAgents
Mixed#26 / 187rank by quality72.8±4.7quality / 1004/7 benchmarks measured
Self-hostingEstimated
How we estimate →What it would cost to run these open weights yourself on rented GPUs, compared with the API price above.
- Hardware
- 1× B200 180GB
- BF16 weights
- Throughput
- ~4,301 tok/s
- many requests batched
- Price
- $0.22–1.93 /M tok
- busy → light use
- On your own machine
- RTX 5090 32GB (int4, ~90 tok/s single-stream)
- single consumer GPU or Mac
Assumptions (5)
- BF16 weights (27.8B params) + 25% KV-cache headroom ≈ 70 GB
- 1× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
- ~4,301 output tok/s aggregate at batch 123 (bandwidth-bound)
- Blended 3:1 input:output; prefill ~19,121 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
~#9 of 21 (estimated)~77.1%
- #1Claude Opus 596.0%
- #2Claude Sonnet 585.2%
- #7Mistral Medium 3.577.6%
- ~#9Qwen3.8 27BEstimated77.1%
- #9Ripple-65B Preview76.5%
No published score — estimated from related benchmarks
SWE-bench Pro (Public)Estimated
~#16 of 24 (estimated)~58.2%
- #1Claude Fable 5.293.0%
- #2Claude Opus 5.589.9%
- #15GPT-5.558.6%
- ~#16Qwen3.8 27BEstimated58.2%
- #16Hy357.9%
No published score — estimated from related benchmarks
DeepSWE v1.1Lab-reported
#40 of 4542.2%
- #1Gemini 4 Argon78.8%
- #2DeepSeek V4.1 Flash 101777.0%
- #39Beam 1.0 501B A23B43.3%
- #40Qwen3.8 27B42.2%
- #41Laguna S 2.140.4%
huggingface.co · 2026-08-14
Terminal-Bench 4.0Verified
#41 of 1015.6%
- #1Kimi K3.1 Pro67.2%
- #2Claude Sonnet 5.563.6%
- #40Muse Spark Open8.1%
- #41Qwen3.8 27B5.6%
- #42Gemini 3.1 Pro4.0%
artificialanalysis.ai · 2026-09-24
AA Coding Agent IndexEstimated
~#23 of 26 (estimated)~36.2
- #1Claude Opus 5.561.7
- #2Claude Opus 559.7
- #22Claude Haiku 5.537.3
- ~#23Qwen3.8 27BEstimated36.2
- #23Gemini 3.8 Flash33.3
No published score — estimated from related benchmarks
Not comparable across labs (1)
Kept for reference, not counted in rankings: these use a lab-specific task set, answer key or scoring, so scores can't be compared fairly across labs. Only published scores are shown.
Writing & chat
IFBenchLab-reported
#11 of 13979.5%
- #1MiniMax-M382.9%
- #2Qwen3.8 2.4T A95B82.8%
- #10Beam 1.0 501B A23B79.6%
- #11Qwen3.8 27B79.5%
- #12Ripple-65B Preview79.2%
huggingface.co · 2026-08-14
#17 of 251671 Elo
eqbench.com · 2026-09-24
Vision
MMMU-ProVerified
#32 of 8776.3%
- #1Claude Opus 5.587.7%
- #2GPT-6 Astra86.9%
- #32GLM-5.3 Flash NVFP476.3%
- #32Qwen3.8 27B76.3%
- #34Ternary Bonsai 2 27B75.5%
artificialanalysis.ai · 2026-09-24
CharXiv ReasoningEstimated
~#7 of 12 (estimated)~80.3%
- #1Gemini 3.8 Flash86.2%
- #2Kimi K384.8%
- #6Nemotron 3.5 Super VL80.8%
- ~#7Qwen3.8 27BEstimated80.3%
- #7Nemotron 3.5 Super VL NVFP479.9%
No published score — estimated from related benchmarks
GDP.pdf (AA)Verified
#22 of 4716.6%
- #1Muse Wonder Full32.6%
- #2GPT-6 Astra31.0%
- #21Grok 4.617.2%
- #22Qwen3.8 27B16.6%
- #23Qwen3.8 Flash-Next15.6%
artificialanalysis.ai · 2026-09-24
ChartographyEstimated
~#16 of 25 (estimated)~24.9%
- #1Claude Fable 5.273.4%
- #2GPT-6 Astra71.0%
- #15Gemini 3.1 Pro26.1%
- ~#16Qwen3.8 27BEstimated24.9%
- #16Grok 4.620.5%
No published score — estimated from related benchmarks
Not comparable across labs (1)
Kept for reference, not counted in rankings: these use a lab-specific task set, answer key or scoring, so scores can't be compared fairly across labs. Only published scores are shown.
Hard reasoning
GPQA DiamondVerified
#30 of 19290.5%
- #1GPT-6 Astra96.1%
- #2Gemini 3.8 Flash95.3%
- #29Beam 1.0 501B A23B90.6%
- #30Qwen3.8 27B90.5%
- #31Qwen3.7 Plus90.0%
artificialanalysis.ai · 2026-09-24
Humanity's Last ExamVerified
#48 of 19833.9%
- #1Claude Fable 5.274.3%
- #2Claude Opus 5.561.4%
- #47Apodex 1.134.1%
- #48Qwen3.8 27B33.9%
- #49LongCat 2.033.7%
artificialanalysis.ai · 2026-09-24
AIME (latest)Estimated
~#4 of 34 (estimated)~95.9%
- #1GLM-5.3 NVFP499.5%
- #2Beam 1.0 501B A23B97.7%
- #3Inkling97.1%
- ~#4Qwen3.8 27BEstimated95.9%
- #4Ternary Bonsai 2 27B95.8%
No published score — estimated from related benchmarks
Agents
τ²-bench (Telecom)Estimated
~#12 of 120 (estimated)~90.9%
- #1Step 3.7 Flash98.5%
- #2DeepSeek-V4-Pro-0813-nvfp4-DSpark98.3%
- #11Ring-2.6-1T92.4%
- ~#12Qwen3.8 27BEstimated90.9%
- #12MiMo-V2.590.6%
No published score — estimated from related benchmarks
OSWorld 2.0Estimated
~#3 of 5 (estimated)~53.6%
No published score — estimated from related benchmarks
BrowseCompEstimated
~#9 of 22 (estimated)~85.0%
- #1Atria Dawn Preview92.5%
- #2GPT-6 Astra91.5%
- #8Gemini 3.1 Pro85.9%
- ~#9Qwen3.8 27BEstimated85.0%
- #9Claude Sonnet 584.7%
No published score — estimated from related benchmarks
GDPval-AA (Elo)Verified
#28 of 911419 Elo
- #1Claude Opus 5.51866
- #2Muse Wonder Full1857
- #27Mistral Large 41422
- #28Qwen3.8 27B1419
- #29K2 Horizon 375B A23B1361
artificialanalysis.ai · 2026-10-10