
Hy4 對決 Kimi K3.1 Pro:兩者皆可下載,僅一款可自行架設
- Orca新OrcaCyber: Hy4 Flash Abliterated / Uncensored2026-11-18$0.07 / $0.28 每百萬 tokens · 182 tok/s
- tencent新Tencent: Hy4 Flash2026-11-16$0.02 / $0.07 每百萬 tokens · 749 tok/s
- Alibaba新Qwen: Qwen4 Max2026-11-16$1.50 / $4.50 每百萬 tokens · 1902 tok/s
- Google新Google: Gemini 4 Carbon2026-11-10$2.00 / $10.00 每百萬 tokens · 87 tok/s
- Anthropic新Anthropic: Claude Fable 5.22026-11-09$8.00 / $40.00 每百萬 tokens · 149 tok/s
- OrcaOrcaCyber: GLM 5.3 Abliterated / Uncensored2026-11-04$3.00 / $5.00 每百萬 tokens · 96 tok/s
- OrcaOrcaCyber: Hy4 Abliterated / Uncensored2026-11-04$0.80 / $1.60 每百萬 tokens · 134 tok/s
- OrcaOrcaCyber: Kimi K3.1 Pro Abliterated / Uncensored2026-11-04$2.00 / $5.00 每百萬 tokens · 83 tok/s
- tencentTencent: Hy42026-11-04$0.83 / $2.50 每百萬 tokens · 118 tok/s
- OrcaOrcaCyber: MiniMax M3.1 Flash Abliterated / Uncensored2026-10-29$0.80 / $1.60 每百萬 tokens · 109 tok/s
- OrcaOrcaCyber: GLM 5.3 Flash Abliterated / Uncensored2026-10-27$0.80 / $1.60 每百萬 tokens · 67 tok/s
- OrcaOrcaCyber: Qwen3.8 Flash Next Abliterated / Uncensored2026-10-26$0.80 / $1.60 每百萬 tokens · 90 tok/s
- OrcaOrcaCyber: MiMo V2.6 Flash Abliterated / Uncensored2026-10-26$0.80 / $1.60 每百萬 tokens · 41 tok/s
- MoonshotMoonshotAI: Kimi K3.1 Pro2026-10-2447智能
- OrcaOrcaCyber: DeepSeek V4.1 Flash Abliterated / Uncensored2026-10-23$0.80 / $1.60 每百萬 tokens · 211 tok/s
- minimaxMiniMax: M3.1 Flash2026-10-1929智能
- OrcaOrca: OrcaCyber Ultra 1.02026-10-17$5.00 / $10.00 每百萬 tokens · 109 tok/s
- deepseekDeepSeek: DeepSeek V4.1 Flash 10172026-10-1733智能
- DeepSeekDeepSeek: DeepSeek V4.1 Flash 09102026-10-17$0.15 / $0.60 每百萬 tokens · 160 tok/s
- DeepSeekDeepSeek: DeepSeek V4.1 Flash2026-10-1740智能
Kimi K3.1 Pro scores 46.5 on the Artificial Analysis Intelligence Index. Tencent Hy4 scores 34.6. That is a 12-point gap, and it is the largest deficit Hy4 runs in any matchup this week — larger than the one against Qwen4-Max, more than ten times the one against Grok 4.7. On the board, Kimi K3.1 Pro is the best open-weight model listed, first of 32 in its class.
Then you look at what the two weigh. Kimi K3.1 Pro is a 2.78-trillion-parameter model with 104 billion active per token. Tencent Hy4 is 774 billion total with 49 billion active. Both were published as downloadable weights — Kimi's on 2 November 2026, Hy4's on 18 November, three weeks apart. Only one of them fits on a machine a normal company owns, and that single fact reorganises the entire comparison. This is not a story about which model is smarter. It is a story about which model you can actually run, and what the smarter one costs you per completed task if you rent it instead.
唯一能決定這件事的基準
以廠商自身建議的精度,達成最小可部署佔用空間:
• Tencent Hy4 — 774 billion parameters in BF16, published as 245 shards of roughly 12.9 GB and 6.4 GB, about 1.45 TB before any quantisation. Tencent's own serving recipe uses tensor parallelism of 8 on a single node's worth of GPUs with the FP8 build, which is roughly 786 GB of weights.
• Kimi K3.1 Pro — 2.78 兆個參數,隨附的儲存庫中有 2.72 TB 的 U8 量化權重,BF16 張量還另外佔用 57 GB。沒有任何單節點配置能讓這個模型成為週末專案。
Both vendors ship the software to make serving tractable, and both are unusually good about it. Tencent publishes a vLLM image tag (vllm/vllm-openai:hy4) and an SGLang image (lmsysorg/sglang:hy4) with the tensor-parallel and speculative-decoding flags written out; Kimi's repository ships a compressed-tensors MXFP4 configuration and a 4-bit path that brings the weight file down to the 2.72 TB figure above. Neither of those is a small-model story. The difference is that an 8-GPU node is a purchase most engineering organisations can authorise, and a 2.8-trillion-parameter deployment is a data-centre decision.
如果自我架設是你對開放權重感興趣的原因,那這句話就是完整的論據,而且它不依賴任何基準測試。

十二個點所在之處
12 分的指數差距聽起來很平均。其實不然——Hy4 在多項評測上表現不落下風,卻在其中一項上被徹底擊敗。
• Agentic coding, Terminal-Bench 4.0 — Kimi K3.1 Pro 67.2%, Tencent Hy4 37.4%. A 30-point gap, and the single biggest contributor to the index difference.
• Agentic science in a terminal, Terminal-Bench-Science — Kimi 34.3%, Hy4 12.9%.
• Software engineering in real repositories, FrontierCode-AA 2.0 — Hy4 31.4%, Kimi 25.5%. Hy4 is the better model here, by six points, and it is the one evaluation where a terminal agent's real-repository work favours Tencent.
• Hard reasoning, HLE-Diamond — Kimi 37.3%, Hy4 32.7%. Closer than the headline suggests.
• Physics, CritPt v2 hard subset — Kimi 49.5%, Hy4 39.0%.
• SaaS workflow execution, AutomationBench-AA — Kimi 73.3%, Hy4 65.6%.
• Agentic knowledge work, AA-Briefcase Elo — Kimi 1,673.6, Hy4 1,445.9. A 228-Elo gap with non-overlapping intervals.
The FrontierCode result is the one worth pausing on, because it contradicts the obvious reading of the other six. On a benchmark of changing real open-source repositories, Hy4 leads. On a benchmark of a clean terminal environment, Kimi leads by 30 points. Tencent's model card explains why that could be: Hy4 was trained specifically against Tencent's own game and frontend pipelines — Unreal, Unity, Godot, UI state management — and reports a 19.1% gain on its internal GameDev suite over Hy4 preview. That is a model tuned for building things, evaluated on a test of repairing them. Take both rows and the honest conclusion is that the coding story is workload-dependent, not that one model codes better.

以每項完成的任務計算,價格低 3.6 倍
這正是比較從爭論轉為算術的地方。Artificial Analysis 報告了兩款模型在每項 Intelligence Index 任務上的實測成本,其中納入了 token 價格、推理長度與快取行為:
• Tencent Hy4 — $1.03 per task, at $0.834 input and $2.501 output per million tokens, 95% cache discount at $0.042.
• Kimi K3.1 Pro — 每項任務 $3.68,每百萬個 token 輸入 $2.00、輸出 $10.00,快取享 90% 折扣,價格為 $0.20。
So the 12-point index gap costs 3.6 times as much per task. Per index point, Kimi is roughly 4.4x the price. That is not a scandal — a model that is genuinely better at hard agentic work is supposed to cost more — but it reframes the decision. If your workload needs 67% on Terminal-Bench, there is no substitute and you should pay. If your workload needs 37%, paying three and a half times for the other 30 points is money spent on capability you will not invoke.
There is a second-order cost that the per-task figure hides. Kimi takes 39.7 minutes per index task against Hy4's 14.0, and returns its first answer token after 162.8 seconds against 115.9. At 30 output tokens per second against Hy4's 93, the slowest model in this comparison by a factor of three is also the most expensive. In an agent loop that runs hundreds of steps, latency and cost compound together rather than trading off.
兩份授權條款,各有不同的觸發陷阱
這兩個模型都屬於開放權重,意思是你可下載並使用它們。兩者都不是 Apache 2.0,而它們商業條款之間的差異並非學術性的:
• Tencent Hy4 — the Tencent Hy License 1.0, derived from Apache 2.0. Section 10 requires a separate written agreement before offering the model as a service if your worldwide gross revenue exceeds $50 million in any 12-month period, or if the product carrying it exceeds 100 million monthly active users. Internal use is carved out.
• Kimi K3.1 Pro — Kimi K3.1-Pro License。限制形式相同,但營收門檻為 2,000 萬美元,而非 5,000 萬美元。
Both carve out end-user products where the model is embedded in specific features rather than exposed as general-purpose model access. Both include mere relaying to a third-party model service as permitted. In plain terms: a small product team building a feature on either model is fine. A company running an inference business on either needs a phone call, and Kimi's phone call comes at less than half the revenue.
為求完整起見,上一代是個例外:於 2026 年 8 月 28 日推出的 Hy4 預覽版採用 Apache 2.0,完全沒有商業門檻。任何依據該預覽版授權進行開發、如今轉向 Hy4 的人,都是在轉向不同的條款。
兩者都執行,尚未選擇
這項比較之所以不必只在紙上定論,是因為這兩個模型都能從同一個端點取得 ,並按供應商自身的費率原樣傳遞: tencent/hy4 為每百萬個詞元 $0.834 與 $2.501, kimi/kimi-k3.1-pro 為 $2.00 與 $10.00,快取讀取為 $0.20。那些就是上述數字,未加價,而供應商價格變動會在同一天反映在這裡,而不是等到續約時。
For a pair this far apart on capability and cost, the useful pattern is not a split — it is a ceiling. Route to kimi/kimi-k3.1-pro through the routing DSL, set tencent/hy4 as the automatic failover target, and watch how often the failover fires. If it fires rarely, you are paying for Kimi's capability and getting it. If it fires constantly because of Kimi's latency, you have just discovered something a benchmark table cannot tell you, at Hy4's prices rather than Kimi's. Model fusion is the other option: a panel of the two would let the stronger model's reasoning carry the weaker model's cheaper token budget on the evaluations where they converge.

什麼會改變這個答案
Three things, in order of likelihood. First, a Hy4 fine-tune: the weights have been public for two days, and the 30-point Terminal-Bench gap is exactly the kind of margin a targeted fine-tune on a company's own terminal traces erodes. Second, an independent evaluation of Kimi K3.1 Pro at a higher reasoning effort than the default max the index uses — the 943,718-token output ceiling in the model's record has not been exercised by anyone publishing results. Third, a Hy4 Flash-class sibling that inherits the agentic training at a quarter of the price, which would make the cost argument against Kimi even harder to answer.
Until one of those lands, the split is clean. Kimi K3.1 Pro is the better model and the more expensive one, by a factor that holds up whether you measure per token, per task or per index point. Tencent Hy4 is the one you can put on your own hardware, at a licence threshold twice as generous and a price a third as steep. Neither of those is a reason to regret the other.
本文中的比較2
根據本文內容識別 · 基準測試:Artificial Analysis · 每日更新
