Skip to content
Model Pareto

Bonsai 27B

PrismMLOpen weights1-bit (Q1_0_g128)huggingface.co ↗
Input
—
Output
—
Speed
~84tok/sest.
Context
262Ktokens

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× 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

~#6 of 21 (estimated)~79.0%
  1. #1Claude Opus 596.0%
  2. #2Claude Sonnet 585.2%
  3. #5MiniMax-M380.5%
  4. ~#6Bonsai 27BEstimated79.0%
  5. #6Hy378.0%
No published score — estimated from related benchmarks
~#10 of 24 (estimated)~63.3%
  1. #1Claude Fable 5.293.0%
  2. #2Claude Opus 5.589.9%
  3. #9GPT-5.6 Terra63.4%
  4. ~#10Bonsai 27BEstimated63.3%
  5. #10Claude Sonnet 563.2%
No published score — estimated from related benchmarks
DeepSWE v1.1Estimated
~#43 of 46 (estimated)~32.5%
  1. #1Gemini 4 Argon78.8%
  2. #2DeepSeek V4.1 Flash 101777.0%
  3. #42Nex-N2.5-Mini36.1%
  4. ~#43Bonsai 27BEstimated32.5%
  5. #43Hy328.0%
No published score — estimated from related benchmarks
~#24 of 102 (estimated)~27.5%
  1. #1Kimi K3.1 Pro67.2%
  2. #2Claude Sonnet 5.563.6%
  3. #22GLM-5.3 Flash32.8%
  4. ~#24Bonsai 27BEstimated27.5%
  5. #24DeepSeek V4.1 Flash26.8%
No published score — estimated from related benchmarks
SciCodeEstimated
~#43 of 119 (estimated)~51.0%
  1. #1Claude Fable 5.267.7%
  2. #2Claude Opus 5.566.9%
  3. #42MiMo-V2.6-Flash51.3%
  4. ~#43Bonsai 27BEstimated51.0%
  5. #43DeepSeek V4 Pro (0813)51.0%
No published score — estimated from related benchmarks
~#13 of 26 (estimated)~47.9
  1. #1Claude Opus 5.561.7
  2. #2Claude Opus 559.7
  3. #12Mistral Large 448.3
  4. ~#13Bonsai 27BEstimated47.9
  5. #13DeepSeek V4.1 Flash47.3
No published score — estimated from related benchmarks

Writing & chat

Vision

MMMUEstimated
~#2 of 3 (estimated)~71.3%
  1. #1Claude Haiku 4.573.2%
  2. ~#2Bonsai 27BEstimated71.3%
  3. #2Llama 4 Maverick69.4%
No published score — estimated from related benchmarks
~#10 of 12 (estimated)~77.1%
  1. #1Gemini 3.8 Flash86.2%
  2. #2Kimi K384.8%
  3. #9Inkling78.1%
  4. ~#10Bonsai 27BEstimated77.1%
  5. #10Claude Sonnet 577.0%
No published score — estimated from related benchmarks
~#5 of 8 (estimated)~82.7%
  1. #1MiniMax-M391.6%
  2. #2Kimi K391.1%
  3. #4Gemini 3.1 Pro85.3%
  4. ~#5Bonsai 27BEstimated82.7%
  5. #5Claude Haiku 4.579.6%
No published score — estimated from related benchmarks
GDP.pdf (AA)Estimated
~#33 of 48 (estimated)~10.0%
  1. #1Muse Wonder Full32.6%
  2. #2GPT-6 Astra31.0%
  3. #31Qwen3.7 Plus12.2%
  4. ~#33Bonsai 27BEstimated10.0%
  5. #33Muse Glimmer10.0%
No published score — estimated from related benchmarks
ChartographyEstimated
~#21 of 25 (estimated)~13.7%
  1. #1Claude Fable 5.273.4%
  2. #2GPT-6 Astra71.0%
  3. #20Grok 4.714.7%
  4. ~#21Bonsai 27BEstimated13.7%
  5. #21DeepSeek V4.1 Flash11.6%
No published score — estimated from related benchmarks

Hard reasoning