ZMIME
Comparison · 3 models · Updated Nov 2, 2026

Kimi K2 Instruct vs Claude Haiku 4.5 vs Gemini 2.5 Flash

Gemini 2.5 Flash comes out ahead, 66 to 57 and 50 on our weighted score, and it is the cheaper option too.

  1. Moonshot AI

    Kimi K2 Instruct

    Released Jul 11, 2025

    50/100
    • ECI139.9
    • Price$0.57 / $2.30
    • Context131K
    Full specs Remove Kimi K2 Instruct

  2. Anthropic

    Claude Haiku 4.5

    Released Oct 15, 2025

    57/100
    • ECI142.2
    • Price$1.00 / $5.00
    • Context200K
    Full specs Remove Claude Haiku 4.5
  3. Our pick

    Google

    Gemini 2.5 Flash

    Released Jun 17, 2025

    66/100
    • ECI140.6
    • Price$0.30 / $2.50
    • Context1.05M
    Full specs Remove Gemini 2.5 Flash
01 — Verdict

Gemini 2.5 Flash is our pick

Gemini 2.5 Flash is the better all-round choice, scoring 66/100 against Claude Haiku 4.5 (57) and Kimi K2 Instruct (50). It leads on price, inputs & features and context window. Claude Haiku 4.5 wins on capability. The score weighs capability 50%, price 25%, inputs & features 15%, context window 10%.

  • CapabilityClaude Haiku 4.5Capabilities Index (ECI): Claude Haiku 4.5 142.2 · Gemini 2.5 Flash 140.6 · Kimi K2 Instruct 139.9
  • Lowest priceGemini 2.5 FlashGemini 2.5 Flash $0.85 · Kimi K2 Instruct $1.00 · Claude Haiku 4.5 $2.00 per 1M tokens (3:1 blend)
  • Longest contextGemini 2.5 FlashGemini 2.5 Flash 1,048,576 · Claude Haiku 4.5 200,000 · Kimi K2 Instruct 131,072 tokens
  • Widest inputsGemini 2.5 FlashKimi K2 Instruct: Text · Claude Haiku 4.5: Text, Images, PDFs · Gemini 2.5 Flash: Text, Images, PDFs, Audio, Video
  • Self-hostingKimi K2 InstructPublishes downloadable weights (Modified MIT)
How the score is built
MeasureWeightKimi K2 InstructClaude Haiku 4.5Gemini 2.5 Flash
CapabilityCapabilities Index (ECI)50%646765
Price25%483451
Inputs & features15%2580100
Context window10%253261
Overall100%50/10057/10066/100
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 Claude Haiku 4.5 vs Gemini 2.5 Flash specifications side by side
SpecificationKimi K2 InstructMoonshot AIClaude Haiku 4.5AnthropicGemini 2.5 FlashGoogle
Capability
Capabilities Index (ECI)139.9142.2 (best)140.6
ECI rank#112 of 165#102 of 165 (best)#110 of 165
GPQA DiamondGraduate-level science questions—71.2%—
OTIS Mock AIME 2024–2025Competition mathematics—66.7%—
SimpleQA VerifiedShort factual questions—13.2%—
Price per million tokens
Input$0.57$1.00$0.30 (best)
Output$2.30 (best)$5.00$2.50
Cached input—$0.10$0.03 (best)
Blended (3:1)$1.00$2.00$0.85 (best)
Long-context rateSame rateSame rateSame rate
Price sourceMedian of 11 providersOfficial Anthropic APIOfficial Google API
Limits
Context window131,072 tokens200,000 tokens1,048,576 tokens (best)
Max output32,768 tokens64,000 tokens65,536 tokens (best)
Inputs and features
TextYesYesYes
ImagesNoYesYes
PDFsNoYesYes
AudioNoNoYes
VideoNoNoYes
ReasoningNoYesYes
Tool callingYesYesYes
Structured outputNoYesYes
Availability
WeightsOpenModified MITProprietaryProprietary
API model ID—claude-haiku-4-5gemini-2.5-flash
API providers1148 (best)32
ReleasedJul 11, 2025Oct 15, 2025Jun 17, 2025
Knowledge cutoff—Feb 28, 2025Jan 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.

  • Kimi K2 Instruct$10.30
  • Claude Haiku 4.5$20.00
  • Gemini 2.5 Flash$8.00
04 — Questions

Which should you choose?

Which is better: Kimi K2 Instruct, Claude Haiku 4.5 or Gemini 2.5 Flash?

Gemini 2.5 Flash is the better all-round choice, scoring 66/100 against Claude Haiku 4.5 (57) and Kimi K2 Instruct (50). It leads on price, inputs & features and context window. Claude Haiku 4.5 wins on capability. The score weighs capability 50%, price 25%, inputs & features 15%, context window 10%.

Which is cheaper, Kimi K2 Instruct, Claude Haiku 4.5 or Gemini 2.5 Flash?

Gemini 2.5 Flash is cheaper at $0.30 input / $2.50 output per million tokens (official Google API price). Kimi K2 Instruct costs $0.57 input / $2.30 output per million tokens (median across 11 API providers); Claude Haiku 4.5 costs $1.00 input / $5.00 output per million tokens (official Anthropic API price). At a typical mix of three input tokens to one output token, that is $0.85 per million tokens for Gemini 2.5 Flash versus $1.00 for Kimi K2 Instruct (1.2× as much) and $2.00 for Claude Haiku 4.5 (2.4× as much).

Which scores higher on benchmarks?

Claude Haiku 4.5 scores higher on the Capabilities Index (ECI): Claude Haiku 4.5 142.2 (#102 of 165), Gemini 2.5 Flash 140.6 (#110 of 165) and Kimi K2 Instruct 139.9 (#112 of 165). The confidence ranges of the top two overlap (139.4–143.9 vs 138.4–142.1), so treat the gap as small.

Which is better for coding?

There are no published SWE-bench Verified results for Kimi K2 Instruct, Claude Haiku 4.5 and Gemini 2.5 Flash yet, so there is no like-for-like coding score. On overall capability, Claude Haiku 4.5 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?

Gemini 2.5 Flash has the largest context window at 1,048,576 tokens, against 200,000 for Claude Haiku 4.5 and 131,072 for Kimi K2 Instruct. Maximum output per response: Kimi K2 Instruct up to 32,768, Claude Haiku 4.5 up to 64,000, Gemini 2.5 Flash up to 65,536 tokens.

Which can read images, PDFs, audio or video?

Kimi K2 Instruct accepts text; Claude Haiku 4.5 accepts text, images and PDFs; Gemini 2.5 Flash accepts text, images, PDFs, audio and video. Gemini 2.5 Flash handles the widest range of inputs.

Are any of these open source?

Kimi K2 Instruct publishes its weights (Modified MIT) and can be self-hosted; Claude Haiku 4.5 and Gemini 2.5 Flash is proprietary.

Which is newer?

Claude Haiku 4.5 is the newest, released Oct 15, 2025. Kimi K2 Instruct came out Jul 11, 2025; Gemini 2.5 Flash came out Jun 17, 2025. Knowledge cutoff: Claude Haiku 4.5 Feb 28, 2025, Gemini 2.5 Flash Jan 2025.

How do you decide the winner?

Each model gets a 0–100 score on capability (50%, independent benchmark results); price (25%, blended price per million tokens (3 input : 1 output), log scale); inputs & features (15%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (10%, 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 2, 2026.

05 — Keep comparing

Related comparisons