Skip to content
Model Pareto

Maple-Preview

DeepGroveOpen weightsnative ternaryhuggingface.co ↗
Compare with nearest rivalsCompare with…
Input
—
Output
—
Speed
~130tok/sest.
Context
131Ktokens

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× H200 141GB
native ternary weights
Throughput
~16,365 tok/s
many requests batched
Price
$0.023–0.19 /M tok
busy → light use
On your own machine
RTX 5090 32GB (native ternary, 300+ tok/s single-stream)
single consumer GPU or Mac
Assumptions (5)
  • NATIVE TERNARY weights (20.2B params, 1.49B active per token) + 25% KV-cache headroom ≈ 5 GB
  • 1× H200 141GB at $3.59–$7.91/GPU-hour on-demand (2026-09-24)
  • ~16,365 output tok/s aggregate at batch 256 (bandwidth-bound); MoE compute scales with active params
  • Blended 3:1 input:output; prefill ~313,784 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
~#5 of 14 (estimated)~80.4%
  1. #1Claude Opus 596.0%
  2. #2Claude Sonnet 585.2%
  3. #4MiniMax-M380.5%
  4. ~#5Maple-PreviewEstimated80.4%
  5. #5Hy378.0%
No published score — estimated from related benchmarks
SWE-bench Pro (Public)Estimated
~#7 of 19 (estimated)~64.5%
  1. #1Claude Opus 5.589.9%
  2. #2Claude Sonnet 5.581.3%
  3. #6Hy4 preview65.7%
  4. ~#7Maple-PreviewEstimated64.5%
  5. #7GPT-5.6 Terra63.4%
No published score — estimated from related benchmarks
DeepSWE v1.1Estimated
~#29 of 32 (estimated)~48.9%
  1. #1GPT-6.1 Sol75.2%
  2. #2DeepSeek V4.1 Flash74.2%
  3. #28Qwen3.8 Max (0902)51.0%
  4. ~#29Maple-PreviewEstimated48.9%
  5. #29Qwen3.8 27B42.2%
No published score — estimated from related benchmarks
Terminal-Bench 4.0Estimated
~#18 of 85 (estimated)~23.2%
  1. #1Claude Sonnet 5.563.6%
  2. #2Claude Opus 5.559.6%
  3. #17Qwen3.8 Flash-Next25.3%
  4. ~#18Maple-PreviewEstimated23.2%
  5. #18MiMo-V2.6-Flash22.7%
No published score — estimated from related benchmarks
SciCodeEstimated
~#32 of 103 (estimated)~50.8%
  1. #1Claude Opus 5.566.9%
  2. #2Claude Fable 5.163.1%
  3. #31DeepSeek V4 Pro (0813)51.0%
  4. ~#32Maple-PreviewEstimated50.8%
  5. #32Qwen3.8 Flash-Next50.6%
No published score — estimated from related benchmarks
AA Coding Agent IndexEstimated
~#9 of 15 (estimated)~52.4
  1. #1Claude Opus 5.566.0
  2. #2Claude Fable 5.162.2
  3. #8GLM-5.353.6
  4. ~#9Maple-PreviewEstimated52.4
  5. #9Kimi K351.9
No published score — estimated from related benchmarks

Hard reasoning

Humanity's Last ExamEstimated
~#65 of 182 (estimated)~20.6%
  1. #1Claude Opus 5.561.4%
  2. #2Claude Fable 5.159.1%
  3. #64Nemotron 3 Super 120B A12B20.8%
  4. ~#65Maple-PreviewEstimated20.6%
  5. #65o320.1%
No published score — estimated from related benchmarks
AA Intelligence IndexEstimated
~#45 of 78 (estimated)~20.6
  1. #1Claude Opus 5.557.6
  2. #2Claude Sonnet 5.556.0
  3. #44K2 Horizon 7B20.6
  4. ~#45Maple-PreviewEstimated20.6
  5. #45Ling 3.0 Flash20.1
No published score — estimated from related benchmarks