FULL STORY
Tencent Hunyuan Hy4 Flash: From Launch to Real-World Tests
Tencent Hunyuan launched the ultra-cheap Hy4 Flash, claiming Haiku 5.5-level performance at ~18x lower cost. Third-party benchmarks and free-trial tests followed, roughly matching DeepSeek V4.1 Flash.
2026-11-16 ~ 2026-11-17 · 3 episodes · 22 posts
Episode 1 · Tencent Hunyuan Launches Hy4 flash, Matching Haiku 5.5 in Blind Tests at ~1/18 the Cost (2026-11-16, 18 posts)
Tencent Hunyuan officially released the cost-efficient model Hy4 flash on November 16, positioning it as the lowest-cost model for everyday productivity. The company says it has 224B total parameters (including 64B Engram parameters) with only 12B activated, supports multimodality and 1M context, and incorporates hybrid attention and other techniques in pretraining; API pricing starts at $0.003 per million tokens for cached input, and the model will be open-sourced soon.
Confirmed
- Official human blind tests show Hy4 flash performs on par with Claude Haiku 5.5 (high) on everyday tasks, at roughly 18x lower cost per task
- In replies to the launch post, the team added cost comparisons: about ¥0.2 per task, 13.8x cheaper than Hy4 (¥5.7) and 24.5x cheaper than Hy4 preview (¥3.2); internal WorkIndex performance is 42.1
- The API is now live, with prices substantially lower than Hy4
Why it matters
- Hy4 flash targets everyday productivity at an extremely low unit price, going directly after Anthropic's lightweight Haiku 5.5; if the blind-test results hold up under third-party verification, it could disrupt high-throughput, cost-sensitive API applications
- @XianbaoQIEN argues the HY4 family and DeepSeek Flash are complementary rather than head-to-head competitors, calling the strategy smart — a sign of Chinese models increasingly specializing into distinct niches
- If the open-source plan materializes, it will further lower the barrier to accessing low-cost models
- Tencent releases Hy4flash: 224B MoE with 12B active, 1/20 the cost of Hy4 — 腾讯混元 · 2026-11-16
- HY4 Flash API launches: ~0.2 RMB per task, beats GPT-6 Luna, open source soon — Xianbao_QIAN · 2026-11-16
- Tencent launches Hunyuan 4flash: 224B-total/12B-active, tasks cost $0.036 — 智东西 · 2026-11-16
- Tencent Hunyuan launches Hy4 flash: matches Claude Haiku 5.5 in blind tests at 1/18 the cost — TencentHunyuan · 2026-11-16
- Tencent Hunyuan launches Hy4 flash: matches Claude Haiku 5.5 at 1/18th the cost — TencentHunyuan · 2026-11-16
- Tencent shares Hy4 flash cost breakdown: ¥0.2 per task, 13.8x cheaper than Hy4 — TencentHunyuan · 2026-11-16
- Tencent launches Hy4 flash: per-task cost just 1/8 of DeepSeek Flash — vista8 · 2026-11-16
- Tencent releases Hy4 Flash at $0.07/$0.281 per M tokens, 24.5x cheaper per task than Hy4 — chrisbrehs · 2026-11-16
- Tencent releases Hunyuan Hy4flash: 1/8 the per-task cost of DeepSeek Flash — AGI Hunt · 2026-11-16
- Tencent Launches Hy4flash: 1/20 the Cost of Hy4 With Faster Agent Execution — APPSO · 2026-11-16
- Hy4 flash unveiled: 224B total params with just 12B active, 1M context, cheaper API — noctrex · 2026-11-16
- Tencent's Hy4 Flash costs $0.07/$0.281 per M tokens, 24.5x cheaper per task than Hy4 — EmetInteractive · 2026-11-16
- Tencent releases Hy4 Flash: 224B total params, 12B active, 1M context — EmetInteractive · 2026-11-16
- Tencent Hunyuan launches Hy4 flash: $0.03 per task, 13.8x cheaper than Hy4 — karthik_r_n · 2026-11-16
- Tencent launches Hy4 Flash: matches Claude Haiku 5.5 at 18x lower cost — TencentHunyuan · 2026-11-17
- Tencent Hunyuan launches Hy4 flash: matches Claude Haiku 5.5 in blind tests at 18x lower task cost — TencentHunyuan · 2026-11-17
- Tencent's hy4 flash: 12B active params, 1M context, ~1/8 DeepSeek cost per task — DammDogg · 2026-11-17
- Tencent's Hy4 Flash lands on kenari with 1M context, image input and tool use — TencentHunyuan · 2026-11-17
Episode 2 · Tencent Hunyuan Hy4 Flash Scores 5.74 on RuneBench at $0.03 per Run (2026-11-17, 2 posts)
Independent tests show Tencent Hunyuan Hy4 Flash (high reasoning mode) scores 5.74 on RuneBench at roughly $0.03 per run, making it one of the cheapest frontier-level options.
- Tencent Hunyuan Hy4 Flash hits 5.74 on RuneBench at ~$0.03 per run, cheapest on Pareto frontier — mariofilhoml · 2026-11-17
- Tencent Hy4 Flash scores 5.74 on RuneBench at just $0.03 per run, cheapest on Pareto frontier — mariofilhoml · 2026-11-17
Episode 3 · Tencent Hy4 Flash Benchmarked Against DeepSeek V4.1 Flash (2026-11-17, 2 posts)
Users tested Tencent's Hy4 Flash, finding it roughly on par with DeepSeek V4.1 Flash for everyday tasks. In a one-shot comparison it used 4x fewer tokens but produced slightly lower-quality output.
- Hands-on: Tencent Hy4 Flash matches DeepSeek V4.1 Flash level and is free for two weeks — TencentHunyuan · 2026-11-17
- Mechanical dragonfly test: Hy4 Flash uses 4x fewer tokens but DeepSeek build is better — TencentHunyuan · 2026-11-17