Nex-N2.5-Pro
- Input
- $0.075/M tok
- Output
- $0.25/M tok
- Blended
- $0.119/M tok · 3:1
- Speed
- ~90tok/sest.
- Context
- 262Ktokens
- Cost per task
- ~$0.10/taskest.
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, compared with the API price above.
- 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 21 (estimated)~83.6%
No published score — estimated from related benchmarks
SWE-bench Pro (Public)Lab-reported
#11 of 2361.2%
- #1Claude Fable 5.293.0%
- #2Claude Opus 5.589.9%
- #10Claude Sonnet 563.2%
- #11Nex-N2.5-Pro61.2%
- #12Atria Dawn Preview59.6%
github.com · 2026-09-08
DeepSWE v1.1Lab-reported
#36 of 4555.8%
- #1Gemini 4 Argon78.8%
- #2DeepSeek V4.1 Flash 101777.0%
- #35Qwen3.8 2.4T A95B56.6%
- #36Nex-N2.5-Pro55.8%
- #37Qwen3.8 Max (0902)51.0%
github.com · 2026-09-08
Terminal-Bench 4.0Estimated
~#26 of 102 (estimated)~26.6%
- #1Kimi K3.1 Pro67.2%
- #2Claude Sonnet 5.563.6%
- #24Mistral Large 426.8%
- ~#26Nex-N2.5-ProEstimated26.6%
- #26Grok 4.725.8%
No published score — estimated from related benchmarks
LiveCodeBenchEstimated
~#6 of 28 (estimated)~88.9%
- #1Qwen3.8-Omni-Flash92.6%
- #2Qwen3.8 Flash-Next91.9%
- #5Nemotron 3 Ultra89.0%
- ~#6Nex-N2.5-ProEstimated88.9%
- #6Beam 1.0 501B A23B88.3%
No published score — estimated from related benchmarks
SciCodeEstimated
~#23 of 119 (estimated)~55.9%
- #1Claude Fable 5.267.7%
- #2Claude Opus 5.566.9%
- #22Claude Opus 556.4%
- ~#23Nex-N2.5-ProEstimated55.9%
- #23DeepSeek V4.1 Flash 101755.9%
No published score — estimated from related benchmarks
AA Coding Agent IndexEstimated
~#14 of 26 (estimated)~47.1
- #1Claude Opus 5.561.7
- #2Claude Opus 559.7
- #13DeepSeek V4.1 Flash47.3
- ~#14Nex-N2.5-ProEstimated47.1
- #14Grok 4.647.0
No published score — estimated from related benchmarks
Agents
τ²-bench (Telecom)Estimated
~#17 of 120 (estimated)~89.6%
- #1Step 3.7 Flash98.5%
- #2DeepSeek-V4-Pro-0813-nvfp4-DSpark98.3%
- #16Ling-2.6-1T89.8%
- ~#17Nex-N2.5-ProEstimated89.6%
- #17KAT Coder Pro V289.5%
No published score — estimated from related benchmarks
OSWorld 2.0Estimated
~#3 of 5 (estimated)~55.3%
No published score — estimated from related benchmarks
BrowseCompLab-reported
#5 of 2189.7%
- #1Atria Dawn Preview92.5%
- #2GPT-6 Astra91.5%
- #4Claude Opus 590.8%
- #5Nex-N2.5-Pro89.7%
- #6Step 5 Preview88.7%
github.com · 2026-09-08
GDPval-AA (Elo)Estimated
~#21 of 92 (estimated)~1512 Elo
- #1Claude Opus 5.51866
- #2Muse Wonder Full1857
- #20Kimi K31533
- ~#21Nex-N2.5-ProEstimated1512
- #21GPT-6 Sol1487
No published score — estimated from related benchmarks
AutomationBench-AAEstimated
~#29 of 85 (estimated)~43.2%
- #1Claude Fable 5.276.4%
- #2Muse Wonder Full71.3%
- #28GPT-5.547.3%
- ~#29Nex-N2.5-ProEstimated43.2%
- #29K2 Horizon 375B A23B37.2%
No published score — estimated from related benchmarks
τ-Bench Banking (AA)Estimated
~#20 of 91 (estimated)~34.3%
- #1Muse Spark 1.350.5%
- #2GLM-5.350.3%
- #19Ling-3.0-flash-VL34.4%
- ~#20Nex-N2.5-ProEstimated34.3%
- #20K2 Horizon 375B A23B34.2%
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.