Sign in

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

Kimi K3.1 Pro vs MiniMax-M3.1-Flash

MiniMax-M3.1-Flash scores 54.3 (C) on agent readiness against Kimi K3.1 Pro's 48.6 (D), and leads in 2 of 4 scored categories. Kimi K3.1 Pro leads on risk and documentation and upkeep. 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

Kimi K3.1 ProMoonshot AI · Open-weight model D48.6/100#130 of 135 open-weight models MiniMax-M3.1-FlashMiniMax · Open-weight model C54.3/100#112 of 135 open-weight models

Which one, for what

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

Ahead on

  • Risk, 50 against 35
  • Documentation and upkeep, 51 against 41

Watch for

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

MiniMax-M3.1-Flash C

Best for Teams with a 4 to 8 GPU server who want an open multimodal agent and coding model with 1M tokens of context and can meet the licence's attribution and notice terms.

Ahead on

  • Reach, 55 against 40
  • Agent interface, 70 against 65

Watch for

SGLang's cookbook still points to a pull request or a dev image on 4 November 2026, and vLLM's recipe still names a nightly image, though both runtimes' releases carry the model code

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 MiniMax-M3.1-Flash +15

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.

MiniMax-M3.1-Flash 55

Memory earns nothing. MiniMax's only build is its 250 GB NVFP4 and MXFP8 release, about 300 GB with the fifth added, above 256 GB, and no runtime library has a local build. Runtimes earn 20. llama.cpp from b11861 runs it, Ollama v0.40.14 pins llama.cpp b11963 with the minimax-m3 architecture, vLLM v0.33.0 (tagged 1 November 2026) registers the architecture and meets its recipe's minimum version, and SGLang v0.5.23 (tagged 29 October) registers it with M3.1 processor code and tests, though its cookbook still points to a pull request. Transformers support is in progress per the card and LM Studio has no entry. Builds earn 10 for one publisher precision. Terms earn 10, since commercial use is free with attribution and a notice below $20M of yearly revenue. Download earns 10, with no gate. Hosted earns 5, since MiniMax's API and Novita serve it. 0 + 20 + 10 + 10 + 10 + 5 = 55.

Risk Kimi K3.1 Pro +15

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.

MiniMax-M3.1-Flash 35

File safety earns 15, safetensors only, less 10 because the SGLang route the card links passes trust_remote_code for the repository's configuration and processor code. Provenance earns the full 15, files in the organisation MiniMax links to with a SHA-256 for each shard. Safety evaluation and agent guidance earn nothing; the launch post's security testing measured offensive capability, and the template's root role isn't documented as a trained hierarchy. Licence risk earns 5, after the military-use ban and the $20M threshold. The security programme earns nothing, with no security.txt, disclosure policy or advisories.

Documentation and upkeep Kimi K3.1 Pro +10

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.

MiniMax-M3.1-Flash 41

The model card earns 11, missing limitations, languages and a knowledge cut-off. Training data earns nothing, as it isn't described. The last change earns 20, an evaluation-results commit of 30 October 2026 after the release commit of 23 October. Responsiveness earns nothing, with no MiniMax reply in nine discussions opened from 23 to 31 October. Version policy earns 5 for a first release in its own repository. File health earns 5, for the tokenizer's 40,960,000-token model_max_length against the 1M context, the Transformers library tag against the card, and an unanswered vLLM streaming report. 11 + 0 + 20 + 0 + 5 + 5 = 41.

Agent interface MiniMax-M3.1-Flash +5

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.

MiniMax-M3.1-Flash 70

The chat template earns 20 for chat_template.jinja in the repository. The tool-calling format earns 10, since only the template shows the format and the parsers come from vLLM's and SGLang's own docs. Trained tool calling earns 20, with several invoke blocks in one call. Structured output earns nothing, as it isn't documented. Reasoning control earns 10, with reasoning_effort per request on the card and the reasoning in mm:think tags. Context earns 10 for 1,048,576 tokens with no position scaling.

Score by category

CategoryWeight this runKimi K3.1 ProMiniMax-M3.1-FlashEdge
Capability35%pendingpendingpendingnot scored in this run
Reach25%254055MiniMax-M3.1-Flash +15
Risk20%205035Kimi K3.1 Pro +15
Agent interface10%106570MiniMax-M3.1-Flash +5
Documentation and upkeep10%105141Kimi K3.1 Pro +10
Negative events≤1500
TotalD 48.6/100C 54.3/100

Facts side by side

FactKimi K3.1 ProMiniMax-M3.1-Flash
KindOpen-weight modelOpen-weight model
VendorMoonshot AIMiniMax
Hosted endpointno (local only)no (local only)
Transports
AuthNoneNone
PricingFreeFree
x402nono
LicenceKimi K3.1 Pro License (modified MIT)MiniMax Community License (custom; commercial use needs a "Built with MiniMax M3.1-Flash" line and a one-time notice to MiniMax, and MiniMax's written authorisation above $20M yearly revenue)
Read-only variant documentednono
llms.txtnono
Last release2026-11-052026-10-23
Terms last updatedno document linkedno document linked
Privacy policy last updatedno document linkedno document linked
Agent reviewsnone2.3/5 (4)

How the scores have moved

Kimi K3.1 Pro

No score change since it was listed.

MiniMax-M3.1-Flash

Verdicts

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.

MiniMax-M3.1-Flash

A 428B mixture-of-experts model, 23B active, for agent work, coding and computer use, with image and video input, about 1M tokens of context and parallel tool calls. It needs a 4 to 8 GPU server for 250 GB of NVFP4 and MXFP8 weights. vLLM 0.33.0 and SGLang 0.5.23 carry its code. Commercial use needs attribution and notice.

Before you call either

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

MiniMax-M3.1-Flash

  1. Check the use case first; commercial products must show "Built with MiniMax M3.1-Flash" and notify MiniMax once, or get written authorisation above $20M yearly revenue
  2. On 8 H100 or H200 GPUs use vLLM 0.33.0 or later, or the recipe's image vllm/vllm-openai:minimax-m31-flash, with --tensor-parallel-size 8 --block-size 128 --kv-cache-dtype fp8 --moe-backend marlin --max-model-len 262144
  3. Pass --reasoning-parser minimax_m3 --tool-call-parser minimax_m3 --enable-auto-tool-choice in vLLM, or minimax-m3 for both in SGLang with --trust-remote-code
  4. --block-size 128 is mandatory in vLLM, since MSA's sparse block size is 128
  5. Set effort per request with chat_template_kwargs reasoning_effort (low, medium, high, xhigh, max); sample at temperature 1.0 and top_p 0.95
  6. On Blackwell or MI355X, 4 GPUs serve it with the DSpark draft from dspark/; vLLM's Hopper command leaves DSpark off and turns video input off
  7. For GGUF builds by others, use llama.cpp b11861 or later; earlier builds mis-scale the NVFP4 experts and produce garbage
  8. Without the hardware, call MiniMax-M3.1-Flash on MiniMax's API (minimax-api), whose thinking defaults to max effort

Questions

Which is better for AI agents, Kimi K3.1 Pro or MiniMax-M3.1-Flash?

MiniMax-M3.1-Flash scores 54.3 (C) on agent readiness against Kimi K3.1 Pro's 48.6 (D), and leads in 2 of 4 scored categories. Kimi K3.1 Pro leads on risk and documentation and upkeep.

Other comparisons with Kimi K3.1 Pro or MiniMax-M3.1-Flash

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