Kimi-K3 vs Hy4 preview
Why this pair: Current Tencent and Moonshot AI frontier language models
Benchmark Performance
Available Benchmarks
Technical Differences
Side-by-Side Facts
| Field | Kimi-K3 | Hy4 preview |
|---|---|---|
| Developer | Moonshot AI | Tencent |
| Family | Kimi K3 | Hy4 |
| Model | Kimi-K3 | Hy4 preview |
| Version | Kimi-K3 | Hy4 preview |
| Lifecycle | active | preview |
| Released | 2026-07-16 | 2026-08-28 |
| Knowledge cutoff | Unknown | Unknown |
| Input modalities | Text, Image | Text |
| Output modalities | Text | Text |
| Context window | 1,048,576 | 1,000,000 |
| Total parameters | 2,779,931,837,184 | 770,000,000,000 |
| Active parameters | 104,000,000,000 | 49,000,000,000 |
| License | other | apache-2.0 |
| Open weights | Yes | Yes |
| API available | Yes | No |
| Self-hostable | Yes | Yes |
| Provider access | Deepinfra (Standard), Openrouter (Standard) | Openrouter (Standard) |
| Capabilities | chat, generation, reasoning | chat, generation, reasoning, tools |
17 comparable fields · 14 material differences · Editorially curated pair passes the primary-source comparison gate
Kimi-K3 Capabilities
Hy4 preview Capabilities
Internal Comparison Graph
Related Comparisons
| A | Pair | B | Context |
|---|---|---|---|
Claude Fable 5.1Anthropic | vs | Kimi-K3Moonshot AI | cross-developer peersimage, text |
Claude Mythos 5.1Anthropic | vs | Kimi-K3Moonshot AI | cross-developer peersimage, text |
Hy3Tencent | vs | Hy4 previewTencent | family variantstext |
Claude Fable 5.1Anthropic | vs | Hy4 previewTencent | cross-developer peerstext |
 GPT-5.6 SolOpenAI | vs | Hy4 previewTencent | cross-developer peerstext |
 Gemini 3.1 ProGoogle DeepMind | vs | Hy4 previewTencent | cross-developer peerstext |
 DeepSeek-V4-ProDeepSeek | vs | Hy4 previewTencent | cross-developer peerstext |
| vs | Hy4 previewTencent | cross-developer peerstext | |
Hy4 previewTencent | vs |  GLM-5.3Z.ai | cross-developer peerstext |
 MiniMax-M3MiniMax | vs | Hy4 previewTencent | cross-developer peerstext |
| vs | Hy4 previewTencent | cross-developer peerstext | |
 Mistral Large 3Mistral AI | vs | Hy4 previewTencent | cross-developer peerstext |
Primary Evidence
Sources and Freshness
Questions
Kimi-K3 vs Hy4 preview FAQs
Is Kimi-K3 or Hy4 preview better for coding?+
This comparison does not currently contain a protocol-matched coding benchmark for both Kimi-K3 and Hy4 preview, so Model Markets cannot name a coding leader from pricing, context size, or capability labels alone.
Which is cheaper, Kimi-K3 or Hy4 preview?+
Kimi-K3 is $2.85 and Hy4 preview is $0.834 per million tokens, so Hy4 preview is cheaper on this metric. Kimi-K3 is $14.25 and Hy4 preview is $2.501 per million tokens, so Hy4 preview is cheaper on this metric.
Which has a larger context window, Kimi-K3 or Hy4 preview?+
Kimi-K3 has the larger sourced context window. Kimi-K3 supports 1,048,576 and Hy4 preview supports 1,000,000.
Which performs better in benchmarks, Kimi-K3 or Hy4 preview?+
There is no overall benchmark winner: At least two independently verified, protocol-matched benchmarks are required for an overall winner.
Can Kimi-K3 or Hy4 preview be self-hosted?+
Both models have the same recorded self-hosting status: supported. Kimi-K3 is open weight; Hy4 preview is open weight.
Can Kimi-K3 and Hy4 preview understand images?+
Kimi-K3 is documented with image input; Hy4 preview is not documented with image input. This reflects supported input modalities, not vision quality.
Which can generate longer answers, Kimi-K3 or Hy4 preview?+
Neither has a larger sourced maximum output. Kimi-K3 is — and Hy4 preview is —.
Do Kimi-K3 and Hy4 preview support reasoning and tool use?+
Kimi-K3: reasoning and image input. Hy4 preview: reasoning and tool calling. Feature support does not establish relative quality.
Which is available from more inference providers, Kimi-K3 or Hy4 preview?+
Kimi-K3 has 2 sourced provider routes; Hy4 preview has 1, so Kimi-K3 has broader tracked availability.
Which offers better value, Kimi-K3 or Hy4 preview?+
There is no universal value winner. Compare the input and output prices above with the matched benchmark result for your workload: cheaper tokens can be offset by different quality, token usage, latency, or provider availability.