On 6 July 2026, Tencent announced Hy3. Tencent converted its April preview into a production-focused open release using feedback from more than 50 internal products.
This release brief was checked against first-party material on 10 August 2026. The date above is the public announcement date, not the date a repository was created or a third-party provider added the model. Where access or weights arrived later, that distinction is recorded below.
Release record
| Field | Verified detail |
|---|---|
| Announcement | 6 July 2026 |
| Availability or weight release | 6 July 2026 on Hugging Face and ModelScope, with TokenHub API access |
| Release type | open-weight general-purpose agent model |
| Access | Apache 2.0 BF16 and FP8 weights plus hosted API access |
| Architecture | a 295B-total, 21B-active MoE with 192 experts and a 3.8B MTP layer |
| Maximum stated context | 256K tokens |
What changed
Hy3’s release evidence emphasises reliability improvements after real product use. Tencent reports lower hallucination and commonsense-error rates, lower multi-turn intent-tracking issues, and less than four percentage points of SWE-Bench variance across three agent scaffolds. Those measures are closer to integration risk than a single best-case leaderboard result and help explain why the final release matters beyond the April preview.
The practical comparison is therefore not simply whether Hy3 has the largest headline score. Teams need to ask whether its architecture, access terms, latency, tool behaviour and evaluation setup match the workload they actually intend to run. A model can lead one harness while losing on cost, refusal behaviour, multilingual quality or repeatability in another.
Benchmarks worth retaining
| Evaluation | Reported result | How to read it |
|---|---|---|
| Expert work-task rating | 2.67/4 versus 2.51 for GLM-5.1 | Blind evaluation with 270 experts |
| Hallucination rate | 5.4%, down from 12.5% | Tencent internal production-scenario evaluation |
| Multi-turn issue rate | 7.9%, down from 17.4% | Internal context-retention evaluation |
These are release-time results, not independently reproduced guarantees. Several reliability figures and the 270-expert work-task evaluation are internal; public benchmark charts should still be read with their harness appendix. Scores should remain attached to the disclosed effort setting, agent harness, tool access, timeout, context-management policy and judge model. Moving a number into a procurement sheet without those conditions creates false comparability.
Architecture and access
Hy3 is described as a 295B-total, 21B-active MoE with 192 experts and a 3.8B MTP layer with 256K tokens of stated context. Its access position at verification time is Apache 2.0 BF16 and FP8 weights plus hosted API access. That wording matters: open weights, source-available weights, an API, a product preview and a research demonstration give adopters very different rights and different levels of reproducibility.
Before deployment, record the exact model identifier or checkpoint, inference stack, quantisation, reasoning setting, region, price schedule and supplier terms. If the release uses a custom licence, read the licence itself rather than relying on the word “open” in launch copy. If it is API-only, preserve the dated documentation and change-notice route because the served snapshot can change without a downloadable artefact.
What an evaluation should test next
For Hy3, a credible internal gate should include:
- a frozen set of representative tasks with pass, fail and abstain criteria;
- a matched baseline using the same tools, timeout, prompt budget and reviewer rubric;
- repeated runs to expose variance rather than reporting a single best attempt;
- latency, token use and total task cost alongside task success;
- adversarial, multilingual and long-context cases relevant to the real deployment; and
- rollback evidence showing the previous model can be restored safely.
The wider model change-control guide explains how to keep model, prompt, tool and corpus changes reconstructable. The AI dependency inventory guide covers the release and supplier records needed after deployment.
AIEngine verdict
Hy3 is a strong open production candidate, especially for Chinese and enterprise workflows. Its Apache licence and integration evidence are advantages, but the internal reliability claims should be reproduced on local tasks.
This is a launch assessment, not a certification. Benchmark leadership is useful evidence of where to test; it is not authorization to place the model in a high-impact workflow without domain evaluation, security review and an accountable owner.
Primary sources
Image provenance
Hero image: Tencent official release artwork. The locally served WebP is a crop of the first-party release or model-card asset recorded in the repository provenance manifest.
