Bonsai 27B
- Input
- —
- Output
- —
- Speed
- ~84tok/sest.
- Context
- 262Ktokens
Rankings
Coding
Lab-reported#51 / 137rank by quality57.3±10.4quality / 1001/7 benchmarks measuredWriting & chat
Lab-reported#63 / 162rank by quality56.1±9.7quality / 1001/5 benchmarks measuredVision
Lab-reported#64 / 98rank by quality46.0±8.7quality / 1001/6 benchmarks measuredHard reasoning
Lab-reported#70 / 218rank by quality59.6±9.9quality / 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
- 1-bit (Q1_0_g128) weights
- Throughput
- ~6,454 tok/s
- many requests batched
- Price
- $0.13–1.15 /M tok
- busy → light use
- On your own machine
- RTX 5090 32GB (1-bit (Q1_0_g128), 300+ tok/s single-stream)
- single consumer GPU or Mac
Assumptions (5)
- 1-BIT (Q1_0_G128) weights (27.3B params) + 25% KV-cache headroom ≈ 5 GB
- 1× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
- ~6,454 output tok/s aggregate at batch 184 (bandwidth-bound)
- Blended 3:1 input:output; prefill ~38,942 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)~79.0%
No published score — estimated from related benchmarks
SWE-bench Pro (Public)Estimated
~#10 of 24 (estimated)~63.3%
- #1Claude Fable 5.293.0%
- #2Claude Opus 5.589.9%
- #9GPT-5.6 Terra63.4%
- ~#10Bonsai 27BEstimated63.3%
- #10Claude Sonnet 563.2%
No published score — estimated from related benchmarks
DeepSWE v1.1Estimated
~#43 of 46 (estimated)~32.5%
- #1Gemini 4 Argon78.8%
- #2DeepSeek V4.1 Flash 101777.0%
- #42Nex-N2.5-Mini36.1%
- ~#43Bonsai 27BEstimated32.5%
- #43Hy328.0%
No published score — estimated from related benchmarks
Terminal-Bench 4.0Estimated
~#24 of 102 (estimated)~27.5%
- #1Kimi K3.1 Pro67.2%
- #2Claude Sonnet 5.563.6%
- #22GLM-5.3 Flash32.8%
- ~#24Bonsai 27BEstimated27.5%
- #24DeepSeek V4.1 Flash26.8%
No published score — estimated from related benchmarks
LiveCodeBenchLab-reported
#13 of 2776.4%
- #1Qwen3.8-Omni-Flash92.6%
- #2Qwen3.8 Flash-Next91.9%
- #12gpt-oss-20b77.7%
- #13Bonsai 27B76.4%
- #14Maple-Preview75.1%
huggingface.co · 2026-09-30
SciCodeEstimated
~#43 of 119 (estimated)~51.0%
- #1Claude Fable 5.267.7%
- #2Claude Opus 5.566.9%
- #42MiMo-V2.6-Flash51.3%
- ~#43Bonsai 27BEstimated51.0%
- #43DeepSeek V4 Pro (0813)51.0%
No published score — estimated from related benchmarks
AA Coding Agent IndexEstimated
~#13 of 26 (estimated)~47.9
- #1Claude Opus 5.561.7
- #2Claude Opus 559.7
- #12Mistral Large 448.3
- ~#13Bonsai 27BEstimated47.9
- #13DeepSeek V4.1 Flash47.3
No published score — estimated from related benchmarks
Writing & chat
IFBenchLab-reported
#71 of 13952.4%
- #1MiniMax-M382.9%
- #2Qwen3.8 2.4T A95B82.8%
- #70Jamba Reasoning 3B52.4%
- #71Bonsai 27B52.4%
- #72Qwen3.5 Omni Plus51.2%
huggingface.co · 2026-09-30
LMArena Text, Non-English (Elo)Estimated
~#35 of 57 (estimated)~1419 Elo
- #1Gemini 4 ArgonFew votes1512
- #2Claude Opus 5.51497
- #34Gemma 4 26B A4BFew votes1423
- ~#35Bonsai 27BEstimated1419
- #35Muse GlimmerFew votes1418
No published score — estimated from related benchmarks
LMArena Text (Elo)Estimated
~#36 of 57 (estimated)~1434 Elo
- #1Gemini 4 Argon1525
- #2Claude Opus 5.51507
- #35MiMo-V2.51434
- ~#36Bonsai 27BEstimated1434
- #36Mistral Medium 3.51426
No published score — estimated from related benchmarks
EQ-Bench Creative Writing v3 (Elo)Estimated
~#9 of 26 (estimated)~1917 Elo
No published score — estimated from related benchmarks
Vision
MMMU-ProLab-reported
#63 of 8760.5%
- #1Claude Opus 5.587.7%
- #2GPT-6 Astra86.9%
- #62Nova 2.0 Omni61.9%
- #63Bonsai 27B60.5%
- #64Claude Haiku 4.558.6%
huggingface.co · 2026-09-30
CharXiv ReasoningEstimated
~#10 of 12 (estimated)~77.1%
- #1Gemini 3.8 Flash86.2%
- #2Kimi K384.8%
- #9Inkling78.1%
- ~#10Bonsai 27BEstimated77.1%
- #10Claude Sonnet 577.0%
No published score — estimated from related benchmarks
OmniDocBench v1.5Estimated
~#5 of 8 (estimated)~82.7%
- #1MiniMax-M391.6%
- #2Kimi K391.1%
- #4Gemini 3.1 Pro85.3%
- ~#5Bonsai 27BEstimated82.7%
- #5Claude Haiku 4.579.6%
No published score — estimated from related benchmarks
GDP.pdf (AA)Estimated
~#33 of 48 (estimated)~10.0%
- #1Muse Wonder Full32.6%
- #2GPT-6 Astra31.0%
- #31Qwen3.7 Plus12.2%
- ~#33Bonsai 27BEstimated10.0%
- #33Muse Glimmer10.0%
No published score — estimated from related benchmarks
ChartographyEstimated
~#21 of 25 (estimated)~13.7%
- #1Claude Fable 5.273.4%
- #2GPT-6 Astra71.0%
- #20Grok 4.714.7%
- ~#21Bonsai 27BEstimated13.7%
- #21DeepSeek V4.1 Flash11.6%
No published score — estimated from related benchmarks
Hard reasoning
GPQA DiamondEstimated
~#78 of 193 (estimated)~80.6%
- #1GPT-6 Astra96.1%
- #2Gemini 3.8 Flash95.3%
- #77Qwen3.5 9B80.6%
- ~#78Bonsai 27BEstimated80.6%
- #78G9v3-39A5B80.5%
No published score — estimated from related benchmarks
Humanity's Last ExamEstimated
~#64 of 199 (estimated)~27.3%
- #1Claude Fable 5.274.3%
- #2Claude Opus 5.561.4%
- #63GPT-5.4 mini28.1%
- ~#64Bonsai 27BEstimated27.3%
- #64MiMo-V2.527.2%
No published score — estimated from related benchmarks
AIME (latest)Lab-reported
#17 of 3387.1%
- #1GLM-5.3 NVFP499.5%
- #2Beam 1.0 501B A23B97.7%
- #15Ternary Bonsai 27B87.5%
- #17Bonsai 27B87.1%
- #18Mistral Medium 3.586.3%
huggingface.co · 2026-09-30
AA Intelligence IndexEstimated
~#39 of 103 (estimated)~29.7
- #1Gemini 4 Carbon59.4
- #2Claude Opus 5.557.6
- #38K2 Horizon 375B A23B30.5
- ~#39Bonsai 27BEstimated29.7
- #39Gemini 3.1 Pro29.7
No published score — estimated from related benchmarks