Hunyuan-A13B
Open 80B-total/13B-active MoE with dual-mode fast/slow reasoning.
About
Tencent's first open-source MoE: 80B total parameters with only 13B active per token, pretrained on a filtered 20T-token corpus with extra STEM curation. Its signature feature is dual-mode chain-of-thought — a fast-thinking mode for routine queries and a slow-thinking mode for multi-step problems — plus native FP8/Int4 quantization support for cheap inference.
The license is Tencent's own open-source terms rather than a standard OSI license, so read the repo LICENSE before commercial deployment. Benchmark claims are Tencent-reported.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
- On par with o1 and DeepSeek on mainstream benchmarks (Tencent-reported)
- Pretrained on 20T filtered tokens with STEM-focused curation
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (verified tier, metrics-v1). Evidence: Official Tencent-Hunyuan GitHub repo, released 2025-06-27; 80B total / 13B active MoE, 256K context, dual-mode CoT from tech report; GitHub license field shows NOASSERTION — custom Tencent open terms, not OSI-standard; Benchmarks are Tencent-reported. License: Custom Tencent open-source license; GitHub shows NOASSERTION. Read the repo LICENSE before commercial use..
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 7.0/10
- Openness · license + docs
- 6.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · Official Tencent-Hunyuan GitHub repo, released 2025-06-27
- · 80B total / 13B active MoE, 256K context, dual-mode CoT from tech report
- · GitHub license field shows NOASSERTION — custom Tencent open terms, not OSI-standard
- · Benchmarks are Tencent-reported