TencentHy4
As of 29 Oct 2026, Hy4 by Tencent scores 55.7 on the BenchLeader Index (581 ranked configurations), with its best results in maths (66) and weakest in agents & tools (49). It costs $1.25 per 1M tokens blended ($0.834 in / $2.50 out), and streams about 47 tokens/s (OpenRouter). Released 28 Aug 2026; 6 independent results so far.
Key facts (as of 29 Oct 2026)
- Hy4 costs $1.25 per 1M tokens blended ($0.834 in / $2.50 out) (list price, 29 Oct 2026).
- Hy4 streams about 47 tokens/s (OpenRouter, 29 Oct 2026).
- Hy4 was released on 28 Aug 2026 (Tencent).
In detail. Hy4 is a Tencent open-weights reasoning model, released 28 Aug 2026. Its best configuration ranks #194 of 581 on the BenchLeader Index at 55.7 ±6.3, in the upper half. The ± is the point: 313 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in maths (66) and lowest in agents & tools (49). At $1.25 per million tokens blended it is mid-priced. Output speed of 47 tokens per second puts it in the slowest quarter, with a first token in 2.9 s. Last measured 29 Oct 2026.
- Blended price
- $1.25/M
- $0.834 in · $2.50 out
- Output speed
- 47 tok/s
- OpenRouter traffic, 7-day median; not yet measured by Artificial Analysis
- First answer
- 2.93 s
- Context
- 1.0M
- Full answer
- –
- Cost per run
- –
- Released
- 28 Aug 2026
- Overall index56
- Coding61
- Agents & tools49
- Maths66
- Knowledge58
How its index has moved
27 Oct 2026 to 29 Oct 2026The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Coding
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena WebDev | 1631#22 | Arena (LMArena) | |
| IOI | 59.3%#18 | Vals AI | |
| Code Migrationnot in index | 47.4%#14 | Vals AI | |
| Code Migration v1.1not in index | 45.4%#19 | Vals AI | |
| ProgramBenchnot in index | 0.0%#32 | Vals AI | |
| Vibe Code Bench v1.1not in index | 77.5%#36 | Vals AI |
Agents & tools
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Agent | 1.2#22 | Arena (LMArena) | |
| Terminal-Bench 2.1 (Vals) | 55.1%#48 | Vals AI | |
| CyberBenchnot in index | 65.4%#31 | Vals AI | |
| Finance Agent v2not in index | 55.1%#24 | Vals AI | |
| Finance Agent v2.1not in index | 61.6%#16 | Vals AI | |
| Harvey's Legal Agent Benchmark v1.1not in index | 8.8%#21 | Vals AI | |
| Harvey's Legal Agent Benchmark v1not in index | 9.2%#22 | Vals AI | |
| Legal Research Benchnot in index | 45.2%#19 | Vals AI | |
| SREBenchnot in index | 2.3%#24 | Vals AI | |
| Tax Agent Benchnot in index | 32.6%#19 | Vals AI | |
| Terminal-Bench 4.0 (Vals)not in index | 8.1%#45 | Vals AI | |
| Terminal-Bench Sciencenot in index | 1.4%#36 | Vals AI |
Maths
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| ProofBench | 75.0%#16 | Vals AI |
Knowledge
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| LegalBench | 83.8%#48 | Vals AI | |
| BioMysteryBenchnot in index | 69.3%#14 | Vals AI | |
| Excel Modeling Benchmarknot in index | 58.1%#39 | Vals AI | |
| Excel Modeling Benchmark v1.1not in index | 69.0%#21 | Vals AI | |
| MedCodenot in index | 43.3%#47 | Vals AI | |
| MedScribenot in index | 83.6%#45 | Vals AI | |
| Public Benefits Bench v1.1not in index | 68.6%#9 | Vals AI |
Composite
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Vals Indexnot in index | 53.2#21 | Vals AI | |
| Vals Finance Indexnot in index | 57.5%#14 | Vals AI | |
| Vals Router Indexnot in index | 53.2%#16 | Vals AI |
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| Cohere | 82 tok/s | 1.35 s | $0.834 | $2.50 | 262k | – |
| Tencent Cloud | 46 tok/s | 3.09 s | $0.751 | $2.25 | 1.0M | fp8 |
| GMICloud | 42 tok/s | 1.87 s | $0.834 | $2.50 | 262k | bf16 |
| NovitaAI | 41 tok/s | 3.85 s | $0.834 | $2.50 | 1M | fp8 |
| AtlasCloud | 35 tok/s | 3.62 s | $0.834 | $2.50 | 1.0M | fp8 |
| DeepInfra | 31 tok/s | 1.52 s | $0.792 | $2.38 | 1.0M | fp4 |
| SiliconFlow | 26 tok/s | 1.93 s | $0.834 | $2.50 | 1.0M | fp8 |
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.042 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0011 | $0.0008 | 9.4 s |
| Summarize a 30-page report | 12,000 / 600 | $0.011 | $0.0044 | 15.8 s |
| Code edit | 6,000 / 1,500 | $0.0088 | $0.0052 | 35.2 s |
| Agentic coding session | 60,000 / 4,000 | $0.060 | $0.024 | 1.5 min |
| Structured extraction | 2,000 / 200 | $0.0022 | $0.0010 | 7.2 s |
More from Tencent
See also
Questions people ask
- What are Hy4's benchmark scores?
- As of 29 Oct 2026, Hy4 scores 55.7 ±6.3 on the BenchLeader Index, strongest in maths (66) and weakest in agents & tools (49). That rests on 6 independent benchmark results, and every score on this page links to its publisher.
- How does Hy4 rank?
- Hy4 is #194 of 581 ranked configurations on 29 Oct 2026. 313 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 69.1.
- How much does Hy4 cost per 1M tokens?
- As of 29 Oct 2026, Hy4 costs $0.834 per 1M input tokens and $2.50 per 1M output tokens, $1.25 blended (three input tokens per output token).
- When was Hy4 released?
- Tencent released Hy4 on 28 Aug 2026.
Data as of 29 Oct 2026. Compare with another model.
Cite as: BenchLeader, “Hy4: benchmarks, pricing, speed and rank”, data as of 29 Oct 2026, https://www.benchleader.com/models/hy4.