Qwen3.5 4B
- Input
- $0.03/M tok
- Output
- $0.15/M tok
- Blended
- $0.06/M tok · 3:1
- Speed
- 22tok/s
- Context
- 262Ktokens
Long-tail import from Artificial Analysis (2026-09-24); AA variant: Qwen3.5 4B (Non-reasoning). Price = AA median host price.
Rankings
Writing & chat
Verified#103 / 162rank by quality31.5±9.5quality / 1001/5 benchmarks measuredPrice frontierVision
Verified#71 / 98rank by quality36.6±8.2quality / 1001/6 benchmarks measuredPrice frontierHard reasoning
Verified#138 / 218rank by quality34.1±5.9quality / 1002/4 benchmarks measuredPrice frontierAgents
Verified#97 / 187rank by quality31.0±7.8quality / 1002/7 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
- ~15,768 tok/s
- many requests batched
- Price
- $0.048–0.43 /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)
- BF16 weights (4B params) + 25% KV-cache headroom ≈ 10 GB
- 1× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
- ~15,768 output tok/s aggregate at batch 256 (bandwidth-bound)
- Blended 3:1 input:output; prefill ~132,891 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.
Writing & chat
IFBenchVerified
#105 of 13933.3%
- #1MiniMax-M382.9%
- #2Qwen3.8 2.4T A95B82.8%
- #104EXAONE 4.0 32B33.5%
- #105Qwen3.5 4B33.3%
- #106LFM2.5-VL-1.6B33.1%
artificialanalysis.ai · 2026-09-24
MMMLUEstimated
~last of 5 (estimated)below measured range (<81.3%)
No published score — estimated from related benchmarks
LMArena Text, Non-English (Elo)Estimated
~#44 of 57 (estimated)~1363 Elo
- #1Gemini 4 ArgonFew votes1512
- #2Claude Opus 5.51497
- #43Qwen3 Next 80B A3B Instruct1386
- ~#44Qwen3.5 4BEstimated1363
- #44GLM-4.6VFew votes1361
No published score — estimated from related benchmarks
LMArena Text (Elo)Estimated
~#44 of 57 (estimated)~1382 Elo
- #1Gemini 4 Argon1525
- #2Claude Opus 5.51507
- #43Qwen3 Next 80B A3B Instruct1399
- ~#44Qwen3.5 4BEstimated1382
- #44GLM-4.6VFew votes1378
No published score — estimated from related benchmarks
EQ-Bench Creative Writing v3 (Elo)Estimated
~#19 of 26 (estimated)~1565 Elo
- #1GPT-6 Astra2173
- #2Claude Fable 5.12162
- #18Inkling1611
- ~#19Qwen3.5 4BEstimated1565
- #19Gemini 3.5 Flash-Lite1559
No published score — estimated from related benchmarks
Vision
MMMU-ProVerified
#59 of 8762.1%
- #1Claude Opus 5.587.7%
- #2GPT-6 Astra86.9%
- #59Llama 4 Maverick62.1%
- #59Qwen3.5 4B62.1%
- #61Gemma 4 12B62.0%
artificialanalysis.ai · 2026-09-24
CharXiv ReasoningEstimated
~last of 12 (estimated)below measured range (<74.5%)
No published score — estimated from related benchmarks
OmniDocBench v1.5Estimated
~#6 of 8 (estimated)~78.6%
- #1MiniMax-M391.6%
- #2Kimi K391.1%
- #5Claude Haiku 4.579.6%
- ~#6Qwen3.5 4BEstimated78.6%
- #6Mistral Small 476.4%
No published score — estimated from related benchmarks
GDP.pdf (AA)Estimated
~#38 of 48 (estimated)~5.9%
- #1Muse Wonder Full32.6%
- #2GPT-6 Astra31.0%
- #37Qwen3.6 35B A3B6.6%
- ~#38Qwen3.5 4BEstimated5.9%
- #38Claude Haiku 4.53.8%
No published score — estimated from related benchmarks
ChartographyEstimated
~last of 25 (estimated)below measured range (<9.0%)
No published score — estimated from related benchmarks
Hard reasoning
GPQA DiamondVerified
#112 of 19271.2%
- #1GPT-6 Astra96.1%
- #2Gemini 3.8 Flash95.3%
- #111K2 Think V271.3%
- #112Qwen3.5 4B71.2%
- #113MiniCPM5-2B70.2%
artificialanalysis.ai · 2026-09-24
Humanity's Last ExamVerified
#128 of 1988.0%
- #1Claude Fable 5.274.3%
- #2Claude Opus 5.561.4%
- #127Ling-2.6-1T8.7%
- #128Qwen3.5 4B8.0%
- #129Qwen3 Next 80B A3B Instruct7.6%
artificialanalysis.ai · 2026-09-24
AIME (latest)Estimated
~#23 of 34 (estimated)~64.2%
- #1GLM-5.3 NVFP499.5%
- #2Beam 1.0 501B A23B97.7%
- #22Phi-4-reasoning-plus78.0%
- ~#23Qwen3.5 4BEstimated64.2%
- #23Mistral Large 338.0%
No published score — estimated from related benchmarks
AA Intelligence IndexEstimated
~#74 of 103 (estimated)~13.9
- #1Gemini 4 Carbon59.4
- #2Claude Opus 5.557.6
- #73Mistral Medium 3.514.2
- ~#74Qwen3.5 4BEstimated13.9
- #74Command A+13.1
No published score — estimated from related benchmarks
Agents
τ²-bench (Telecom)Verified
#20 of 11987.7%
- #1Step 3.7 Flash98.5%
- #2DeepSeek-V4-Pro-0813-nvfp4-DSpark98.3%
- #20MiniCPM-V 4.6 1.3B87.7%
- #20Qwen3.5 4B87.7%
- #22HyperCLOVA X SEED Think (32B)87.4%
artificialanalysis.ai · 2026-09-24
OSWorld-VerifiedEstimated
~#12 of 15 (estimated)~59.3%
- #1Kimi K384.8%
- #2Qwen3.8 27B84.3%
- #11Muse Glimmer65.9%
- ~#12Qwen3.5 4BEstimated59.3%
- #12Nemotron 3.5 Super VL NVFP452.3%
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)~55.4%
- #1Atria Dawn Preview92.5%
- #2GPT-6 Astra91.5%
- #18Nemotron 3.5 Bolt 74B-A9B70.0%
- ~#19Qwen3.5 4BEstimated55.4%
- #19Mistral Medium 3.548.6%
No published score — estimated from related benchmarks
GDPval-AA (Elo)Estimated
~#60 of 92 (estimated)~715 Elo
- #1Claude Opus 5.51866
- #2Muse Wonder Full1857
- #59Claude Haiku 4.5719
- ~#60Qwen3.5 4BEstimated715
- #60MiniCPM5-2B694
No published score — estimated from related benchmarks
AutomationBench-AAEstimated
~#55 of 85 (estimated)~4.5%
- #1Claude Fable 5.276.4%
- #2Muse Wonder Full71.3%
- #54Gemma 4 31B4.9%
- ~#55Qwen3.5 4BEstimated4.5%
- #55LongCat 2.04.5%
No published score — estimated from related benchmarks