Skip to content
Model Pareto

Nex-N2.5-Pro

Nex AGIOpen weightshuggingface.co ↗
Compare with nearest rivalsCompare with…
Input
—
Output
—
Speed
~92tok/sest.
Context
262Ktokens

397B-A17B MoE (Qwen3.5-MoE arch, config.json: 512 experts, 10 active/token + shared expert). Weights were marked "coming soon" in the README at the Sep 8 2026 launch but are live on HF now (checked 2026-09-28, last modified 2026-09-17). Ships natively quantized in FP8 (config.json quantization_config: format "float-quantized", FP8_BLOCK).

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
4× B200 180GB
FP8 weights
Throughput
~12,969 tok/s
many requests batched
Price
$0.20–1.76 /M tok
busy → light use
On your own machine
Mac Studio M5 Ultra 512GB (int4, ~99 tok/s single-stream)
single consumer GPU or Mac
Assumptions (5)
  • FP8 weights (397B params, 17B active per token) + 25% KV-cache headroom ≈ 496 GB
  • 4× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
  • ~12,969 output tok/s aggregate at batch 371 (bandwidth-bound); MoE compute scales with active params
  • Blended 3:1 input:output; prefill ~250,147 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
~#3 of 14 (estimated)~83.2%
  1. #1Claude Opus 596.0%
  2. #2Claude Sonnet 585.2%
  3. ~#3Nex-N2.5-ProEstimated83.2%
  4. #3Gemini 3.1 Pro80.6%
No published score — estimated from related benchmarks
Terminal-Bench 4.0Estimated
~#17 of 84 (estimated)~24.5%
  1. #1Claude Sonnet 5.563.6%
  2. #2Claude Opus 5.559.6%
  3. #16Qwen3.8 Flash-Next25.3%
  4. ~#17Nex-N2.5-ProEstimated24.5%
  5. #17MiMo-V2.6-Flash22.7%
No published score — estimated from related benchmarks
LiveCodeBenchEstimated
~#6 of 20 (estimated)~87.7%
  1. #1Qwen3.8-Omni-Flash92.6%
  2. #2Qwen3.8 Flash-Next91.9%
  3. #5gpt-oss-120b87.8%
  4. ~#6Nex-N2.5-ProEstimated87.7%
  5. #6Gemma 4 31B80.0%
No published score — estimated from related benchmarks
SciCodeEstimated
~#11 of 96 (estimated)~57.5%
  1. #1Claude Opus 5.566.9%
  2. #2Claude Fable 5.163.1%
  3. #10GPT-6 Sol57.6%
  4. ~#11Nex-N2.5-ProEstimated57.5%
  5. #11Grok 4.757.4%
No published score — estimated from related benchmarks
AA Coding Agent IndexEstimated
~#10 of 15 (estimated)~51.6
  1. #1Claude Opus 5.566.0
  2. #2Claude Fable 5.162.2
  3. #9Kimi K351.9
  4. ~#10Nex-N2.5-ProEstimated51.6
  5. #10Grok 4.647.0
No published score — estimated from related benchmarks

Agents

τ²-bench (Telecom)Estimated
~#8 of 116 (estimated)~93.1%
  1. #1Step 3.7 Flash98.5%
  2. #2Gemini 3.1 Pro95.6%
  3. #7Qwen3.5 122B A10B93.6%
  4. ~#8Nex-N2.5-ProEstimated93.1%
  5. #8Qwen3.7 Plus93.0%
No published score — estimated from related benchmarks
OSWorld 2.0Estimated
~#3 of 5 (estimated)~55.3%
  1. #1GLM-5.3 Flash59.1%
  2. #2Kimi K358.3%
  3. ~#3Nex-N2.5-ProEstimated55.3%
  4. #3GPT-5.6 Terra50.2%
No published score — estimated from related benchmarks
GDPval-AA (Elo)Estimated
~#17 of 87 (estimated)~1492 Elo
  1. #1Claude Opus 5.51846
  2. #2Claude Fable 5.11735
  3. #16Kimi K31524
  4. ~#17Nex-N2.5-ProEstimated1492
  5. #17GPT-6 Sol1487
No published score — estimated from related benchmarks
AutomationBench-AAEstimated
~#24 of 80 (estimated)~43.5%
  1. #1Claude Opus 5.569.5%
  2. #2DeepSeek V4.1 Flash68.9%
  3. #23GPT-5.547.3%
  4. ~#24Nex-N2.5-ProEstimated43.5%
  5. #24K2 Horizon 375B A23B37.2%
No published score — estimated from related benchmarks
τ-Bench Banking (AA)Estimated
~#19 of 89 (estimated)~34.7%
  1. #1Muse Spark 1.350.5%
  2. #2GLM-5.350.3%
  3. #18Motif 335.3%
  4. ~#19Nex-N2.5-ProEstimated34.7%
  5. #19Ling-3.0-flash-VL34.4%
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.