ZMIME
Comparison · 3 models · Updated Nov 18, 2026

Apertus 70B vs Qwen3 VL 235B A22B Thinking vs Kimi K2 Instruct 0905

No independent capability benchmark covers every model here yet. The score uses price, inputs & features and context window.

Price in rating: Included · Exclude price

  1. Swiss AI

    Apertus 70B

    Released Sep 2, 2025

    34/100
    • ECI—
    • Price$0.64 / $2.67
    • Context66K
  2. Alibaba (Qwen)

    Qwen3 VL 235B A22B Thinking

    Released Sep 23, 2025

    45/100
    • ECI—
    • Price$0.50 / $3.95
    • Context131K
  3. Moonshot AI

    Kimi K2 Instruct 0905

    Released Sep 5, 2025

    39/100
    • ECI—
    • Price$0.60 / $2.50
    • Context262K
01 — Verdict

Capability evidence is incomplete

There is not enough shared capability evidence to name an overall winner. The scores (Qwen3 VL 235B A22B Thinking 45/100, Kimi K2 Instruct 0905 39/100, Apertus 70B 34/100) compare the available specifications and price. The score weighs price 50%, inputs & features 30%, context window 20%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.

  • CapabilityNot enough dataNo independent benchmark covers every model here yet
  • Lowest priceKimi K2 Instruct 0905Kimi K2 Instruct 0905 $1.07 · Apertus 70B $1.15 · Qwen3 VL 235B A22B Thinking $1.36 per 1M tokens (3:1 blend)
  • Longest contextKimi K2 Instruct 0905Kimi K2 Instruct 0905 262,144 · Qwen3 VL 235B A22B Thinking 131,072 · Apertus 70B 65,536 tokens
  • Widest inputsQwen3 VL 235B A22B ThinkingApertus 70B: Text · Qwen3 VL 235B A22B Thinking: Text, Images · Kimi K2 Instruct 0905: Text
  • Self-hostingAll open weightsEvery model here can be downloaded and run on your own hardware
How the score is built
MeasureWeightApertus 70BQwen3 VL 235B A22B ThinkingKimi K2 Instruct 0905
Price50%454246
Inputs & features30%256025
Context window20%223243
Specs & price only100%34/10045/10039/100

Left out because at least one model lacks the data: capability. The remaining weights were rescaled.

02 — Side by side

Every spec in one table

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

Apertus 70B vs Qwen3 VL 235B A22B Thinking vs Kimi K2 Instruct 0905 specifications side by side
SpecificationApertus 70BSwiss AIQwen3 VL 235B A22B ThinkingAlibaba (Qwen)Kimi K2 Instruct 0905Moonshot AI
Capability
Capabilities Index (ECI)———
ECI rank———
Price per million tokens
Input$0.64$0.50 (best)$0.60
Output$2.67$3.95$2.50 (best)
Cached input———
Blended (3:1)$1.15$1.36$1.07 (best)
Long-context rateSame rateSame rateSame rate
Price sourceMedian of 2 providersMedian of 5 providersMedian of 13 providers
Limits
Context window65,536 tokens131,072 tokens262,144 tokens (best)
Max output8,192 tokens32,768 tokens98,304 tokens (best)
Inputs and features
TextYesYesYes
ImagesNoYesNo
PDFsNoNoNo
AudioNoNoNo
VideoNoNoNo
ReasoningNoYesNo
Tool callingYesYesYes
Structured outputNoNoNo
Availability
WeightsOpenApache-2.0OpenOpenModified MIT
API model ID———
API providers2514 (best)
ReleasedSep 2, 2025Sep 23, 2025Sep 5, 2025
Knowledge cutoffSep 2025Mar 31, 2025—
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.

  • Apertus 70B$11.74
  • Qwen3 VL 235B A22B Thinking$12.90
  • Kimi K2 Instruct 0905$11.00
04 — Questions

Which should you choose?

Which is better: Apertus 70B, Qwen3 VL 235B A22B Thinking or Kimi K2 Instruct 0905?

There is not enough shared capability evidence to name an overall winner. The scores (Qwen3 VL 235B A22B Thinking 45/100, Kimi K2 Instruct 0905 39/100, Apertus 70B 34/100) compare the available specifications and price. The score weighs price 50%, inputs & features 30%, context window 20%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.

Which is cheaper, Apertus 70B, Qwen3 VL 235B A22B Thinking or Kimi K2 Instruct 0905?

Kimi K2 Instruct 0905 is cheaper at $0.60 input / $2.50 output per million tokens (median across 13 API providers). Apertus 70B costs $0.64 input / $2.67 output per million tokens (median across 2 API providers); Qwen3 VL 235B A22B Thinking costs $0.50 input / $3.95 output per million tokens (median across 5 API providers). At a typical mix of three input tokens to one output token, that is $1.07 per million tokens for Kimi K2 Instruct 0905 versus $1.15 for Apertus 70B (1.1× as much) and $1.36 for Qwen3 VL 235B A22B Thinking (1.3× as much).

Which scores higher on benchmarks?

There is no independent benchmark that covers all three models yet. Apertus 70B has not been scored yet, Qwen3 VL 235B A22B Thinking has not been scored yet and Kimi K2 Instruct 0905 has not been scored yet.

Which is better for coding?

There are no published SWE-bench Verified results for Apertus 70B, Qwen3 VL 235B A22B Thinking and Kimi K2 Instruct 0905 yet, so there is no like-for-like coding score. Test both on a sample of your own tasks. All three support tool calling for agent workflows.

Which has the bigger context window?

Kimi K2 Instruct 0905 has the largest context window at 262,144 tokens, against 131,072 for Qwen3 VL 235B A22B Thinking and 65,536 for Apertus 70B. Maximum output per response: Apertus 70B up to 8,192, Qwen3 VL 235B A22B Thinking up to 32,768, Kimi K2 Instruct 0905 up to 98,304 tokens.

Which can read images, PDFs, audio or video?

Apertus 70B accepts text; Qwen3 VL 235B A22B Thinking accepts text and images; Kimi K2 Instruct 0905 accepts text. Qwen3 VL 235B A22B Thinking handles the widest range of inputs.

Are any of these open source?

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

Which is newer?

Qwen3 VL 235B A22B Thinking is the newest, released Sep 23, 2025. Kimi K2 Instruct 0905 came out Sep 5, 2025; Apertus 70B came out Sep 2, 2025. Knowledge cutoff: Apertus 70B Sep 2025, Qwen3 VL 235B A22B Thinking Mar 31, 2025.

How do you decide the winner?

This rating uses price (50%, blended price per million tokens (3 input : 1 output), log scale); inputs & features (30%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (20%, 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.