ZMIME
Comparison · 3 models · Updated Nov 18, 2026

Kimi K2 Instruct vs MiniMax-M3 vs Qwen3 235B-A22B Thinking 2507

MiniMax-M3 comes out ahead, 73 to 60 and 55 on our weighted score, and it is the cheaper option too.

Price in rating: Excluded · Include price

  1. Moonshot AI

    Kimi K2 Instruct

    Released Jul 11, 2025

    55/100
    • ECI141.3
    • Price$0.57 / $2.30
    • Context131K
  2. Our pick

    MiniMax

    MiniMax-M3

    Released Jun 1, 2026

    73/100
    • ECI147.9
    • Price$0.30 / $1.20
    • Context1.05M
  3. Alibaba (Qwen)

    Qwen3 235B-A22B Thinking 2507

    Released Jul 25, 2025

    60/100
    • ECI143.3
    • Price$0.30 / $2.90
    • Context262K
01 — Verdict

MiniMax-M3 is our pick

MiniMax-M3 is the better all-round choice, scoring 73/100 against Qwen3 235B-A22B Thinking 2507 (60) and Kimi K2 Instruct (55). It leads on capability, inputs & features and context window. The score weighs capability 67%, inputs & features 20%, context window 13%.

  • CapabilityMiniMax-M3Capabilities Index (ECI): MiniMax-M3 147.9 · Qwen3 235B-A22B Thinking 2507 143.3 · Kimi K2 Instruct 141.3
  • Lowest priceMiniMax-M3MiniMax-M3 $0.525 · Qwen3 235B-A22B Thinking 2507 $0.95 · Kimi K2 Instruct $1.00 per 1M tokens (3:1 blend)
  • Longest contextMiniMax-M3MiniMax-M3 1,048,576 · Qwen3 235B-A22B Thinking 2507 262,144 · Kimi K2 Instruct 131,072 tokens
  • Widest inputsMiniMax-M3Kimi K2 Instruct: Text · MiniMax-M3: Text, Images, Video · Qwen3 235B-A22B Thinking 2507: Text
  • Self-hostingAll open weightsEvery model here can be downloaded and run on your own hardware
How the score is built
MeasureWeightKimi K2 InstructMiniMax-M3Qwen3 235B-A22B Thinking 2507
CapabilityCapabilities Index (ECI)67%697571
Inputs & features20%257035
Context window13%326543
Overall100%55/10073/10060/100
02 — Side by side

Every spec in one table

Highlighted cells lead their row. Dashes mean the data is not published.

Kimi K2 Instruct vs MiniMax-M3 vs Qwen3 235B-A22B Thinking 2507 specifications side by side
SpecificationKimi K2 InstructMoonshot AIMiniMax-M3MiniMaxQwen3 235B-A22B Thinking 2507Alibaba (Qwen)
Capability
Capabilities Index (ECI)141.3147.9 (best)143.3
ECI rank#108 of 158#77 of 158 (best)#100 of 158
GPQA DiamondGraduate-level science questions—90.9% (best)80.1%
OTIS Mock AIME 2024–2025Competition mathematics—71.1%86.7% (best)
SimpleQA VerifiedShort factual questions——40.4%
Price per million tokens
Input$0.57$0.30 (best)$0.30 (best)
Output$2.30$1.20 (best)$2.90
Cached input—$0.06—
Blended (3:1)$1.00$0.525 (best)$0.95
Long-context rateSame rateOver 512K: $0.60 / $2.40Same rate
Price sourceMedian of 11 providersOfficial MiniMax (minimax.io) APIMedian of 11 providers
Limits
Context window131,072 tokens1,048,576 tokens (best)262,144 tokens
Max output32,768 tokens512,000 tokens (best)81,920 tokens
Inputs and features
TextYesYesYes
ImagesNoYesNo
PDFsNoNoNo
AudioNoNoNo
VideoNoYesNo
ReasoningNoYesYes
Tool callingYesYesYes
Structured outputNoNoNo
Availability
WeightsOpenModified MITOpenOpenApache 2.0
API model ID—MiniMax-M3—
API providers1152 (best)12
ReleasedJul 11, 2025Jun 1, 2026Jul 25, 2025
Knowledge cutoff———
03 — Cost

What would a month cost?

Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.

  • Kimi K2 Instruct$10.30
  • MiniMax-M3$5.40
  • Qwen3 235B-A22B Thinking 2507$8.80
04 — Questions

Which should you choose?

Which is better: Kimi K2 Instruct, MiniMax-M3 or Qwen3 235B-A22B Thinking 2507?

MiniMax-M3 is the better all-round choice, scoring 73/100 against Qwen3 235B-A22B Thinking 2507 (60) and Kimi K2 Instruct (55). It leads on capability, inputs & features and context window. The score weighs capability 67%, inputs & features 20%, context window 13%.

Which is cheaper, Kimi K2 Instruct, MiniMax-M3 or Qwen3 235B-A22B Thinking 2507?

MiniMax-M3 is cheaper at $0.30 input / $1.20 output per million tokens (official MiniMax (minimax.io) API price). Qwen3 235B-A22B Thinking 2507 costs $0.30 input / $2.90 output per million tokens (median across 11 API providers); Kimi K2 Instruct costs $0.57 input / $2.30 output per million tokens (median across 11 API providers). At a typical mix of three input tokens to one output token, that is $0.525 per million tokens for MiniMax-M3 versus $0.95 for Qwen3 235B-A22B Thinking 2507 (1.8× as much) and $1.00 for Kimi K2 Instruct (1.9× as much).

Which scores higher on benchmarks?

MiniMax-M3 scores higher on the Capabilities Index (ECI): MiniMax-M3 147.9 (#77 of 158), Qwen3 235B-A22B Thinking 2507 143.3 (#100 of 158) and Kimi K2 Instruct 141.3 (#108 of 158). Their confidence ranges do not overlap (145.8–150.0 vs 141.4–145.1), so the gap is a real one.

Which is better for coding?

There are no published SWE-bench Verified results for Kimi K2 Instruct, MiniMax-M3 and Qwen3 235B-A22B Thinking 2507 yet, so there is no like-for-like coding score. On overall capability, MiniMax-M3 leads, which tends to carry over to coding, but test on your own codebase. All three support tool calling for agent workflows.

Which has the bigger context window?

MiniMax-M3 has the largest context window at 1,048,576 tokens, against 262,144 for Qwen3 235B-A22B Thinking 2507 and 131,072 for Kimi K2 Instruct. Maximum output per response: Kimi K2 Instruct up to 32,768, MiniMax-M3 up to 512,000, Qwen3 235B-A22B Thinking 2507 up to 81,920 tokens.

Which can read images, PDFs, audio or video?

Kimi K2 Instruct accepts text; MiniMax-M3 accepts text, images and video; Qwen3 235B-A22B Thinking 2507 accepts text. MiniMax-M3 handles the widest range of inputs.

Are any of these open source?

Yes, all three publish their weights (Modified MIT and Apache 2.0), so you can self-host them.

Which is newer?

MiniMax-M3 is the newest, released Jun 1, 2026. Qwen3 235B-A22B Thinking 2507 came out Jul 25, 2025; Kimi K2 Instruct came out Jul 11, 2025.

How do you decide the winner?

This rating uses capability (67%, independent benchmark results); inputs & features (20%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (13%, maximum tokens per request, log scale). Dimensions missing for any model are dropped and the remaining weights rescaled, so every model is judged on the same evidence. Specs and prices come from public model listings and the labs’ own API pages; capability scores come from independent benchmark runs. Data updated Nov 18, 2026.