Hy4 flash
Launch tracker · day 7
Hy4 flash at launch: the first independent results
As of 22 Nov 2026, Hy4 flash (Tencent, first listed 16 Nov 2026) has independent results from Vals AI (Vals Index 50.8), Design Arena (Design Arena Website 1227) and Terminal-Bench (Terminal-Bench 11.8%). It has no BenchLeader Index yet: the Index needs at least five independent results across two categories. It is listed at $0.035 in / $0.141 out per 1M tokens, with a 1M-token context window. Not yet measured by Artificial Analysis, LMArena, Epoch AI, LiveBench, ARC Prize, Scale AI and SWE-bench.
- First listed
- 16 Nov 2026 (day 7)
- Provider
- Tencent
- Price
- $0.035 in / $0.141 out per 1M tokens
- Context window
- 1M tokens
- Open weights
- Yes: tencent/Hy4-flash
- BenchLeader Index
- Not yet (needs five results in two categories)
| Independent board | Results so far |
|---|---|
| Vals AI | Vals Index 50.8 · Excel Modeling Benchmark v1.1 63.0% · Finance Agent v2.1 60.8% · more below |
| Design Arena | Design Arena Website 1227 · Design Arena Code 1225 |
| Terminal-Bench | Terminal-Bench 11.8% |
Boards usually post results within days of a launch; this page is rebuilt every three hours.
Scores from independent publishers only; benchmark figures a provider publishes about its own model are not shown.
Good value at this level: no ranked model within 2 Index points costs three quarters of the price or less.
No cheaper ranked model scores higher: it is on the price frontier.
As of 22 Nov 2026, Hy4 flash by Tencent is not yet ranked on the BenchLeader Index, which needs at least five independent results across two categories. API pricing: $0.035 per 1M input tokens and $0.141 per 1M output tokens ($0.061 blended at 3:1), list price on 22 Nov 2026. It streams about 69 tokens/s (OpenRouter).
Key facts (as of 22 Nov 2026)
- Hy4 flash API pricing: $0.035 per 1M input tokens and $0.141 per 1M output tokens ($0.061 blended at 3:1) (list price, 22 Nov 2026).
- Hy4 flash streams about 69 tokens/s (OpenRouter, 22 Nov 2026).
In detail. Hy4 flash is a Tencent open-weights reasoning model. It is not yet ranked: the index needs at least five independent results across two categories. At $0.061 per million tokens blended it is among the cheapest fifth of ranked models. Output speed of 69 tokens per second puts it slower than most, with a first token in 2.8 s. It has been measured at 2 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 16 Nov 2026.
- Blended price
- $0.061/M
- $0.035 in · $0.141 out
- via models.dev
- Output speed
- 69 tok/s
- OpenRouter traffic, 7-day median; not yet measured by Artificial Analysis
- First answer
- 2.81 s
- Context
- 1.0M
- via models.dev
- Accepts
- Text · Images
- via models.dev, OpenRouter
- Cost per task
- –
- Tokens per task
- –
- Released
- –
- date not published by our sources
- Agents & tools30
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| no reasoningbest | – | – | 69 tok/s | 2.81 s | $0.0001 | Agents & tools 30 |
| not stated | – | – | 69 tok/s | 2.81 s | $0.0001 |
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 |
|---|---|---|---|
| Vibe Code Bench v1.1not in index | 74.8%#50 | Vals AI | |
| Design Arena Websitenot in index | 1227#54 | Design Arena | |
| Design Arena Codenot in index | 1225#68 | Design Arena |
Agents & tools
| Benchmark | no reasoning | not stated | Source | Trend |
|---|---|---|---|---|
| Terminal-Bench | 11.8%#105 | – | Terminal-Bench | |
| Finance Agent v2.1not in indexeffort not stated | – | 60.8%#25 | Vals AI | |
| Harvey's Legal Agent Benchmark v1.1not in indexeffort not stated | – | 5.4%#33 | Vals AI | |
| Legal Research Benchnot in indexeffort not stated | – | 38.5%#44 | Vals AI | |
| Tax Agent Benchnot in indexeffort not stated | – | 23.3%#52 | Vals AI | |
| Terminal-Bench 4.0 (Vals)not in indexeffort not stated | – | 15.2%#46 | Vals AI |
Knowledge
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Excel Modeling Benchmark v1.1not in index | 63.0%#41 | Vals AI |
Composite
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Vals Indexnot in index | 50.8#35 | Vals AI | |
| Vals Finance Indexnot in index | 52.7%#38 | Vals AI | |
| Vals Legal Indexnot in index | 68.3%#29 | Vals AI |
API pricing: 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 |
|---|---|---|---|---|---|---|
| Tencent Cloud | 65 tok/s | 3.12 s | $0.035 | $0.141 | 1.0M | fp8 |
Usage
How much Hy4 flash is used in practice, as the gateways that publish it report. Usage is not part of the Index. Most used models and labs.
A typical request over the last 30 days: 120.7k tokens per request, 143 input tokens per output token, 95% of input from cache, 58% of output is reasoning, 0.91 tool calls per request.
Apps sending it the most tokens on OpenRouter: Hermes Agent (714B), Nous Research API (104B), OpenHands (65.3B), omp (45.3B), pi (33.2B).
| Provider on Vercel AI Gateway | Throughput | First token | Uptime, 1 day |
|---|---|---|---|
| Tencent | 142 tok/s | 3.63 s | 99.98% |
Source: Usage: OpenRouter rankings · Vercel AI Gateway, CC BY 4.0 · data as of 22 Nov 2026.
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 $0.002 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0001 | $0.0000 | 7.2 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0005 | $0.0002 | 11.5 s |
| Code edit | 6,000 / 1,500 | $0.0004 | $0.0003 | 24.5 s |
| Agentic coding session | 60,000 / 4,000 | $0.0027 | $0.0012 | 1.0 min |
| Structured extraction | 2,000 / 200 | $0.0001 | $0.0000 | 5.7 s |
See also
Questions people ask
- Why does Hy4 flash have no BenchLeader Index score?
- The index needs at least five independent results across at least two categories, and as of 22 Nov 2026 Hy4 flash does not have them yet. Launch-post scores are never used, so it is ranked once independent publishers have measured it.
- How much does Hy4 flash cost per 1M tokens?
- As of 22 Nov 2026, Hy4 flash costs $0.035 per 1M input tokens and $0.141 per 1M output tokens, $0.061 blended (three input tokens per output token).
Data as of 22 Nov 2026. Compare these configurations.
Cite as: BenchLeader, “Hy4 flash: benchmarks, pricing, speed and rank”, data as of 22 Nov 2026, https://www.benchleader.com/models/hy4-flash.