SYLIN

Hy

by Tencent · Model · Language & code models

mixed

Strong and cheap on security tasks, but one user saw broken responses.

58% on AA Cyber Index (#3) and 76% on CWE-Bench-AA (#1) at $0.92 per task

AccessOpen weightsNeeds~229 GBDownloadabout 229 GB for the full Hy4 preview weightsLicenseApache 2.0

Why we say so

Artificial Analysis measured Hy at 58% on its Cyber Index (#3) and 76% on CWE-Bench-AA (#1), at $0.92 per task. An HN commenter using it via OpenRouter reported many broken responses in a basic error-correction test, which is a single voice. Tencent's own claims for Hy4 and Hy4 Flash are the weakest evidence, and Tencent concedes Flash trails Hy4 and DeepSeek on terminal coding.

Promised Nov 8: open weights in the coming days

Versions

VersionReleasedRank
Hy4LATEST
unproven Only Tencent's own claims so far, and the weights are a preview
Aug 28
Hy3Jul 16

Evidence

  1. Nov 17
    Via OpenRouter, produced many broken responses in one user's basic error-correction test
  2. Nov 8
    Tencent Hy4 adds visual understanding and computer use; strong coding, 3D and knowledge work; usable via Tencent Cloud and WorkBuddy
  3. Aug 30
    Tencent opens Hy4 Preview weights: 770B MoE, 49B active, gated DeepSeek sparse attention, Apache 2.0, 1.56TB
  4. Aug 30
    Free through the API for the first 2 weeks; heavy load caused heavy throttling, so the test took over an hour
  5. Aug 30
    Beats Qwen 3.8 Max and DeepSeek V4 Pro, and is close to GLM 5.3 and Kimi K3
Show all 13
  1. Aug 30
    Ties GLM 5.3 Flash on Arena web dev leaderboard
  2. Aug 28
    In Tencent's blind evaluation with 163 experts, Hy4 preview edged out GLM 5.3 and Kimi K3
  3. Aug 28
    Claimed to lead Terminal-Bench 2.1 and DeepSWE-style agentic tests, ahead of GLM 5.3, Kimi K3 and DeepSeek V4 Pro
  4. Aug 28
    Claimed to trade the lead with Claude Opus 5 depending on the task
  5. Aug 28
    Given an unhinted broken Docker app (Flask, Postgres, frontend) via Hermes Agent, it found the root bug and applied the right fix; results verified in the UI
  6. Aug 28
    Architecture adds Identity Hyper Connections (parallel residual lanes) and Gated DSA (selective sparse attention reused across layers) for long context
  7. Jul 22
    Tencent released Hy3 295B MoE in 1-bit GGUF at about 83 GB, served on a single GPU
  8. Jul 22
    Per token it routes 8 of 192 experts per layer (15,168 experts total); about 5 GB of dense attention weights stay in RAM

Cost to run