HU
Huawei
Language model · China ·Jul 2026

openPangu 2.0 Pro

Language Open Source
505B
Parameters
512K
Context
5
Benchmarks

About openPangu 2.0 Pro

openPangu 2.0 Pro is an AI model developed by Huawei, in the language category, released in 2026, made available as an open-source model with 505B params. Huawei's openPangu-2.0-Pro is a 505B/18B MoE language model trained entirely on Ascend NPUs. Features MLA with DSA+SWA hybrid attention and 4-branch mHC architecture. Scores 87.9 on GPQA-Diamond, 95.4 on AIME 2026, 68.5 on SWE-bench Verified, and 65.7 on BrowseComp.

On this page you'll find openPangu 2.0 Pro's full specifications, including a context window of 512K tokens, 5 benchmark results. Review provider pricing and benchmark scores below, or compare openPangu 2.0 Pro head-to-head with other language models.

→ More models by Huawei → All language models → AI benchmarks leaderboard

Run openPangu 2.0 Pro locally — what to buy

Share this build ↗

Sized to what openPangu 2.0 Pro actually needs (~512k context at Q4), not to the biggest GPU: Good = cheapest that runs it, Better = best value, Best = most headroom. Speed is a hardware estimate; anything past ~120 tok/s is shown as instant.

Best
Instinct MI430X
Consumer GPU · price n/a
~63.4 tok/s
Q4 · fits full context · up to 216k ctx
Runs with: ExLlamaV2, KoboldCpp, LM Studio
See full estimate →

Software support

·ExLlamaV2 ·KoboldCpp ·LM Studio ·MLX ·Ollama ·SGLang ·TensorRT-LLM ·llama.cpp ·vLLM ?KTransformers ?text-generation-webui

✓ measured · · compatible · — not supported. Informational only — speed is hardware-based.

Consumer GPU
53.4–63.4 tok/s
4 options →

Community

Would you run openPangu 2.0 Pro again?
No votes yet — be the first.
Did it run? — community reports by hardware

No reports yet. Own this model on some hardware? Be the first to confirm it runs.

Discussion

No comments yet. Start the discussion.

openPangu 2.0 Pro benchmark scores

Compare on the benchmark table →

Scores on standardised evaluations. Higher is better, and the percentile shows where openPangu 2.0 Pro lands among every model we score on that benchmark.

BrowseComp agentic
top 20% of 25 65.7%
LiveCodeBench code
top 10% of 42 85.7%
SWE-bench Verified code
#38 of 56 68.5%
AIME 2026 math
#2 of 3 95.4%
GPQA Diamond reasoning
top 5% of 128 87.9%

Benchmarks only say so much. Have you run openPangu 2.0 Pro yourself? Tell everyone how it went — what it is good at, where it fell over, and on what hardware.

More models like openPangu 2.0 Pro

AI
DMind-1
 AI
COLLECTIVE-1
 AI
Cogito 1 series
 AI
Euler (Euler大模型)
· 2.5B
 AI
SmolLM-1.7B
· 1.7B
 AI
Llama-3-70B-Synthia-v3.5
· 70B

Frequently asked questions

What is openPangu 2.0 Pro? ▶

Huawei's openPangu-2.0-Pro is a 505B/18B MoE language model trained entirely on Ascend NPUs. Features MLA with DSA+SWA hybrid attention and 4-branch mHC architecture. Scores 87.9 on GPQA-Diamond, 95.4 on AIME 2026, 68.5 on SWE-bench Verified, and 65.7 on BrowseComp. It is tracked on GenAIList with its specifications, benchmark scores and provider pricing.

Who created openPangu 2.0 Pro? ▶

openPangu 2.0 Pro was developed by Huawei and released in 2026.

Is openPangu 2.0 Pro open source or proprietary? ▶

openPangu 2.0 Pro is released as open source — its weights are publicly available to download, run locally and fine-tune.

What is openPangu 2.0 Pro's context window? ▶

openPangu 2.0 Pro supports a context window of 512K tokens, and has 505B params.

How much does openPangu 2.0 Pro cost? ▶

Pricing for openPangu 2.0 Pro depends on the provider. See the providers table on this page for the latest API rates.

How does openPangu 2.0 Pro perform on benchmarks? ▶

openPangu 2.0 Pro is benchmarked across 5 evaluations on GenAIList, including BrowseComp (65.7%). See the full benchmark table below and compare it with other models.

No videos yet. Know a good one? Add it below.

Know a great video about openPangu 2.0 Pro?