Sign in

Head to head · Open-weight models you run yourself · October 2026 research run

Hy3 vs Kimi K3.1 Pro

Hy3 scores 62.6 (B) on agent readiness against Kimi K3.1 Pro's 48.6 (D), and leads in every scored category. Both do open-weight models you run yourself.

Best open-weight models you download and run yourself for AI agents · All 4969 open weights comparisons

Hy3Tencent (Hy team) · Open-weight model B62.6/100#28 of 135 open-weight models Kimi K3.1 ProMoonshot AI · Open-weight model D48.6/100#130 of 135 open-weight models

Which one, for what

Hy3 B

Best for Teams with an 8-GPU server who want an agent-tuned MoE model under plain Apache 2.0

Ahead on

  • Reach, 70 against 40
  • Risk, 60 against 50
  • Agent interface, 75 against 65
  • Documentation and upkeep, 56 against 51

Watch for

597.6 GB in BF16, 299.9 GB in FP8 and 168.5 GB at the smallest 4-bit build

Kimi K3.1 Pro D

Best for Operators with a multi-accelerator node who want Moonshot's newest open-weight model for coding and long agent sessions, and everyone else through an API

No category where it leads by five points or more, and no fact that sets it apart.

Watch for

1,561 GB of files, so it needs a node of eight large accelerators at least

Why the edge

Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.

Reach Hy3 +30

Hy3 70

Memory earns 5, because the smallest build of 4-bit precision or better is AngelSlim's GPTQ-Int4 of 168.5 GB, about 202 GB with the fifth added, which is within 256 GB but beyond any single workstation. The 1-bit IQ1_M GGUF fits our reference machine but sits below the 4-bit line. Runtimes earn 20, for model code in released Transformers, vLLM, SGLang and llama.cpp; Ollama and LM Studio have no entry for it. Builds earn the full 15, with Tencent's FP8 repository and AngelSlim's GPTQ and GGUF builds. Terms earn 15 for plain Apache 2.0. Download earns 10, with no gate and no account. Hosted earns 5, since Featherless AI and DeepInfra serve it live through Hugging Face.

Kimi K3.1 Pro 40

Memory 0, the only build is Moonshot's 1,561 GB MXFP4 release, far above 256 GB. Runtimes 5, llama.cpp b12242 alone has the model code in a release, since vLLM's recipe needs 0.34.0 and SGLang's cookbook a branch. Builds 10, one precision. Terms 10, free commercial use below a model-as-a-service revenue threshold. Download 10, not gated. Hosted 5, Moonshot's API and Together serve it. 0 + 5 + 10 + 10 + 10 + 5 = 40.

Risk Hy3 +10

Hy3 60

File safety earns the full 25, safetensors only with no pickle files, and the documented route runs vLLM or SGLang from their own code with a Jinja template, so nothing from the repository runs at load or prompt time. Provenance earns 15, files in the tencent organisation with a SHA-256 for every shard. Safety evaluation earns nothing, since no results from Tencent or anyone else were found. Agent guidance earns nothing, with no cautions, refusal statement or prompt-injection guidance. Licence risk keeps all 15 under plain Apache 2.0, with no territory limit (the EU, UK and South Korea exclusion is in Hy3-preview's licence, not Hy3's). The security programme earns 5 for Tencent's company-wide Security Response Center, which the card doesn't link, and nothing for advisories, since none were found.

Kimi K3.1 Pro 50

File safety 15, safetensors only, less 10 for remote code and prompts built by the repository's script. Provenance 15, Moonshot's own organisation with a hash on every file. Safety evaluation 10, Moonshot's own cyber benchmark results and nobody else's yet. Agent guidance 0, none on the card. Licence risk 10, the revenue threshold. Security programme 0, no way to report a flaw found. 15 + 15 + 10 + 0 + 10 + 0 = 50.

Agent interface Hy3 +10

Hy3 75

The chat template earns 20, a chat_template.jinja in the repository. The tool-calling format earns 10, because the card names the parsers but leaves the format itself to the template and SGLang's cookbook. Trained tool calling earns 20, as the template takes several calls in one block. Structured output earns 5 for the card's claim of stable output formats under constraints, with no word on how to ask for JSON. Reasoning control earns 10, with effort set per request and switchable off, and reasoning returned separately through the named parsers, as the linked vLLM recipe shows. Context earns 10 for 262,144 tokens native.

Kimi K3.1 Pro 65

Chat template 10, the format is only code to run. Tool-calling format 10, the parser is named in the runtimes' recipes and the card doesn't say how tools are declared. Trained tool calling 20, several calls a turn on Kimi K3's tools channel. Structured output 5, named among linked guides without the how. Reasoning control 10, reasoning_effort per request. Context 10, 1M tokens. 10 + 10 + 20 + 5 + 10 + 10 = 65.

Documentation and upkeep Hy3 +5

Hy3 56

The model card earns 11, covering intended use, architecture, parameters and context but not limitations, languages or a knowledge cut-off. Training data earns 5 for a general statement with no cut-off. The last change earns 10, a config.json edit on 20 July and AngelSlim uploads on 21 July, within 180 days on 2 November 2026; the line drops to 5 after 17 January 2027 unless the files change. Responsiveness earns 10, with Tencent replies on 7 of 19 discussions and none since 22 July. Version policy earns the full 10, since the preview stays in its own repository and the card dates it and says what changed. File health earns 10, with files that agree and no broken-file report waiting.

Kimi K3.1 Pro 51

Model card 11, no limitations, languages or cut-off. Training data 5, one general sentence. Last change 20, a commit on 5 November 2026. Responsiveness 0, none of the 8 newest discussions has a reply from Moonshot. Version policy 5, Kimi K3 keeps its repository and no dated note was found. File health 10, the files agree. 11 + 5 + 20 + 0 + 5 + 10 = 51.

Score by category

CategoryWeight this runHy3Kimi K3.1 ProEdge
Capability35%pendingpendingpendingnot scored in this run
Reach25%257040Hy3 +30
Risk20%206050Hy3 +10
Agent interface10%107565Hy3 +10
Documentation and upkeep10%105651Hy3 +5
Negative events≤1500
TotalB 62.6/100D 48.6/100

Facts side by side

FactHy3Kimi K3.1 Pro
KindOpen-weight modelOpen-weight model
VendorTencent (Hy team)Moonshot AI
Hosted endpointno (local only)no (local only)
Transports
AuthNoneNone
PricingFreeFree
x402nono
LicenceApache-2.0Kimi K3.1 Pro License (modified MIT)
Read-only variant documentednono
llms.txtnono
Last release2026-07-202026-11-05
Terms last updatedno document linkedno document linked
Privacy policy last updatedno document linkedno document linked
Agent reviews2.8/5 (4)none

Verdicts

Hy3

Hy3 is a 295B mixture-of-experts model with 21B active, trained for several tool calls a turn, with per-request reasoning effort and 256K of context. It needs an 8-GPU server for FP8 or about 200 GB for a 4-bit build. The main risks are no safety evaluation or agent guidance, and no file changes or replies since July.

Kimi K3.1 Pro

Moonshot's newest flagship, eight days old when read, for operators with a rack. Only llama.cpp has its model code in a release, with vLLM and SGLang support still on a release candidate and a branch. The licence is permissive below a revenue threshold, and Moonshot hasn't answered a discussion on the repository yet.

Before you call either

Hy3

  1. Start it with vLLM on 8 GPUs, vllm serve tencent/Hy3-FP8 --tensor-parallel-size 8 --tool-call-parser hy_v3 --reasoning-parser hy_v3 --enable-auto-tool-choice --served-model-name hy3
  2. In SGLang use --tool-call-parser hunyuan --reasoning-parser hunyuan, as the card shows
  3. Pass reasoning_effort (no_think, low or high) in chat_template_kwargs; the default is no_think, and any other value raises a template error
  4. With the vLLM reasoning parser the chain of thought comes back in message.reasoning and the answer in message.content
  5. Earlier thinking stays in the prompt when tools are present; the template keys this on preserved_thinking, not the interleaved_thinking named in vLLM's recipe
  6. Use temperature 0.9 and top_p 1.0, the card's recommended sampling
  7. On a 128 GB machine only AngelSlim's IQ1_M GGUF fits, llama-server -m Hy3-IQ1_M.gguf -ctk q8_0 -ctv q8_0 -fa on -c 65536, with llama.cpp b11768 or later; at 89.4 GB × 1.2 it sits right at the 107.4 GB ceiling, so shorten the context if it fails to load, and expect a 1-bit build to fall short of Tencent's numbers
  8. Without the hardware, call it through Featherless AI or DeepInfra via Hugging Face's inference providers

Kimi K3.1 Pro

  1. Use Moonshot's API or Together unless you have a node of eight B300-class accelerators
  2. With vLLM use 0.34.0 or later and the recipe's flags, --tool-call-parser kimi_k3 --reasoning-parser kimi_k3
  3. Pass the whole assistant message back in later turns, reasoning_content and tool_calls included, as the card requires
  4. Set reasoning_effort to low, high or max per request. Max is the default and thinking can't be turned off
  5. Kimi K3's model code doesn't load it, because the attention layers changed
  6. Validate JSON output yourself, since the card doesn't document structured output for the weights

Questions

Which is better for AI agents, Hy3 or Kimi K3.1 Pro?

Hy3 scores 62.6 (B) on agent readiness against Kimi K3.1 Pro's 48.6 (D), and leads in every scored category.

Other comparisons with Hy3 or Kimi K3.1 Pro

Machine-readable

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.

Get an auditSee a sample report