Hunyuan HY3-FP8
Model overview
Hy3-FP8
Tencent's FP8-quantized Hy3 instruct checkpoint, tracked as a separate open-weight model surface for local inference, coding, and agent workflows.
general LLMreasoningcodinglong contextagentchatcode generationsoftware engineering agentlocal inference
Representative version
Hy3-FP8
| Company | Tencent / Hunyuan |
|---|---|
| Release | 2026-07-04 |
| Release type | Model checkpoint release |
| Parameters | 295B / 21B active |
| Architecture | FP8-quantized mixture-of-experts transformer |
| Context | 262,144 |
| Input modalities | text / code |
| Output modalities | text / code / tool_call |
| Tool use / function calling | Confirmed: Tool use, Function calling |
| Structured output / reasoning | Confirmed: Reasoning |
| Vision / audio / video | Vision: Not disclosed; Audio: Not disclosed; Video: Not disclosed |
| Open weights / License | Weights available / Apache-2.0 License sourcetencent-hy3-fp8-current · 2026-08-04 |
| Vendor API | Not disclosed API source pending verification |
| OpenRouter | Not separately listed on OpenRouter |
| Pricing | No reliable public source |
| Benchmark | No reliable public source |
| Public signals | 1 runtime / access record / 1 activity snapshot |
| Use cases | complex reasoning / coding and review / Agent workflows / local inference |
| Claim evidence | Release event / Model specifications / Weight checkpointTencent Hy3-FP8 model cardHy3-FP8 · 2026-07-04 · Hunyuan HY3-FP8 · Hy3-FP8 · Hy3-FP8 |
OpenRouter
Pricing / context / availability
| OpenRouter | Not separately listed on OpenRouter |
|---|---|
| OpenRouter listing | Not separately listed in the current OpenRouter public list |
| Mapped version | Hy3-FP8 tencent-hy3-fp8-current |
| Pricing | No reliable public source |
| Context | 262144 |
| Speed/Latency | No reliable public source |
| Snapshot time | 2026-08-11T06:28:09.895Z |
| Source | View source |
Verified limitations
Usage boundaries
- The reviewed official model page does not list a hosted inference provider for Hy3-FP8.
Public activity counters
Hugging Face model-card activity snapshot
| Platform | Entry | Last-month downloads | Current likes | Current Spaces | Time | Source |
|---|---|---|---|---|---|---|
| huggingface Current HF activity snapshot | tencent/Hy3-FP8 | 15,812 | 67 | 0 | View source |
Downloads are Hugging Face's trailing-month counter; likes and Spaces are snapshots at collection time.
Operations
Runtime / access signals
| Time | Signal | Original text | Source |
|---|---|---|---|
| 07/07/2026, 09:06 PM UTC | vllm 0.24 RuntimeError: HYV3ReasoningParser reasoning parser could not locate think start/end tokens in the... | (APIServer pid=143) Traceback (most recent call last): (APIServer pid=143) File "/root/vllm/.venv/bin/vllm", line 10, in <module (APIServer pi... | Source |
Missing public information
Public information not yet available
Not listed on OpenRouterNo reliable benchmark sourceNo linkable model-specific reviewSpeed/latency not disclosed