Pairwise benchmark snapshot · Aug 17, 2026noindex (thin pair)

MoonshotAI: Kimi K2.5 vs StepFun: Step 3.7 Flash benchmark

Name: MoonshotAI: Kimi K2.5 vs StepFun: Step 3.7 Flash AI benchmark comparison
Creator: CompareLLM
License: https://creativecommons.org/licenses/by/4.0/

This page compares MoonshotAI: Kimi K2.5 (Moonshot) and StepFun: Step 3.7 Flash (stepfun) using the latest snapshots we have as of Aug 17, 2026. StepFun: Step 3.7 Flash wins on output price at $1.15/1M. MoonshotAI: Kimi K2.5 has the larger context window (262k tokens). Every cell below is a dated snapshot from a named public source — we do not invent missing scores, and we do not run SWE-bench ourselves.

Share Analysis:

WhatsApp Telegram X / Post LinkedIn Reddit

Moonshot

MoonshotAI: Kimi K2.5

Auto-discovered from OpenRouter (moonshotai/kimi-k2.5). Preview until a second source matches.

K2.5 MaxElo 1,515$2.8499999999999996/1M out

stepfun

StepFun: Step 3.7 Flash

Auto-discovered from OpenRouter (stepfun/step-3.7-flash). Preview until a second source matches.

$1.15/1M out

Model Lineage & Multi-Tier Derivatives🇨🇳 CN Frontier

Kimi Series Family (4 models)

Compare vs Base (Kimi K3)

Kimi K3

Moonshot

Flagship Base

CoT Scaler

Elo

1528

Speed

68 tok/s

Price

$15.00 / 1M

View model →vs this

MoonshotAI: Kimi K2.5

Moonshot

K2.5 Max

CoT Scaler

Elo

1515

Speed

72 tok/s

Price

$2.85 / 1M

-13 Elo-81% cost1.1x speed

Viewing now

MoonshotAI: Kimi K2 0711

Moonshot

K2 Base

Elo

1465

Speed

85 tok/s

Price

$2.30 / 1M

-63 Elo-85% cost1.3x speed

View model →vs this

Kimi Chat 1.5

Moonshot

2M Context

Elo

1420

Speed

95 tok/s

Price

$0.80 / 1M

-108 Elo-95% cost1.4x speed

View model →vs this

MoonshotAI: Kimi K2.5 Inference Compute Profile

Inference-Time Compute & Reasoning EffortDynamic CoT

MoonshotAI: Kimi K2.5 Reasoning Scaler & Token Billing Simulator

Param: thinking_mode

Reasoning models scale test-time compute by generating hidden chain-of-thought (CoT) tokens. These tokens are billed at standard output rates and directly expand latency (TTFT) in exchange for higher GPQA Diamond and SWE-bench Verified problem resolve rates.

Estimated TTFT

270 ms

1.0x (Standard eval)

CoT Thinking Tokens

~2,500

Billed as output tokens

Est. Cost / 1k Calls

$8.27

7.3x base invoice

STEM / SWE Boost

+5.5%

GPQA & SWE-bench scaling

Medium (Deep Reasoning) Operational Profile

Kimi Swarm reasoning trace for long-context planning.

Direct Provider APIMoonshot

"thinking_mode": "medium"

OpenRouter Unified APIUnified Gateway

"reasoning": {
  "effort": "medium",
  "max_tokens": 2500
}

Swap or Add Any Model to this Matchup

Instant pairwise benchmark recalculation across 400+ LLMs

Model A

Model B

Quick vs:vs Claude Opus 4.5 vs Claude Opus 4.6 vs Claude Sonnet 4.5 vs Claude 3.7 Sonnet

Interactive Comparison Diagram

Matchup Breakdown

Tug-of-War Matchup Matrix

Category wins across intelligence, speed, and pricing efficiency.

MoonshotAI: Kimi K2.5 (7)vsStepFun: Step 3.7 Flash (2)

MoonshotAI: Kimi K2.5: 7W (78%)Overall: MoonshotAI: Kimi K2.5StepFun: Step 3.7 Flash: 2W (22%)

← MoonshotAI: Kimi K2.51 tiesStepFun: Step 3.7 Flash →

MoonshotAI: Kimi K2.5 (5/5)

Intelligence & Reasoning

Preference Elo, Coding proficiency, SWE-bench & LiveBench accuracy

Preference Elo

1,515vs—

Coding Elo

1,520vs—

SWE-bench

71.3%vs—

LiveBench

67.2%vs—

GPQA Diamond

87.6%vs—

MoonshotAI: Kimi K2.5: 5WStepFun: Step 3.7 Flash: 0W

MoonshotAI: Kimi K2.5 (2/2)

Speed & Latency

Generation throughput and time to first token responsiveness

Output speed

72 tok/svs—

Time to first token

270 msvs—

MoonshotAI: Kimi K2.5: 2WStepFun: Step 3.7 Flash: 0W

StepFun: Step 3.7 Flash (2/3)

Pricing & Capacity

Cost per million tokens and max context window length

Output price

$2.85/1Mvs$1.15/1M

Input price

$0.57/1Mvs$0.2/1M

Context window

262kvs262k

MoonshotAI: Kimi K2.5: 0WStepFun: Step 3.7 Flash: 2W

Capability Matchup Radar

Multi-Dimensional Capability Radar

Each spoke is a skill. Farther from the center is better (0–100th percentile). Tap any dot to inspect details.

Tap any node to inspect

Percentile 0–100

Dimensional AdvantageMoonshotAI: Kimi K2.5 4/6

Preference Elo

No shared data

Throughput Speed

No shared data

Token Price Efficiency

StepFun: Step 3.7 Flash

$1.70/1M cheaper

Full Benchmark Score Matrix

—

Elo

High is better

1,515

EloHigh is better

—

Code Elo

High is better

1,520

Code EloHigh is better

—

LiveBench

High is better

67.2%

LiveBenchHigh is better

—

SWE-bench

High is better

71.3%

SWE-benchHigh is better

—

GPQA

High is better

87.6%

GPQAHigh is better

—

TTFT

Low is better

270 ms

TTFTLow is better

—

Speed

High is better

72 tok/s

SpeedHigh is better

—

In $

Low is better

$0.57 / 1M

In $Low is better

$0.20 / 1M+185%

Out $

Low is better

$2.85 / 1M

Out $Low is better

$1.15 / 1M+148%

Context

High is better

262k

ContextHigh is better

262k

Interactive Simulator (USD)

Workload Cost & Savings Calculator

Simulate monthly production API costs in USD (US Dollar).

Save up to 61% with StepFun: Step 3.7 Flash

Monthly Volume50M tokens/mo

Quick Presets:

Token Ratio (In vs Out)70% In / 30% Out

RAG / Search (10% out)Coding / Chat (50% out)

MoonshotAI: Kimi K2.5$62.70 / mo

In: $19.95Out: $42.75

StepFun: Step 3.7 Flash$24.25 / mo

In: $7Out: $17.25

Estimated Cost Delta

StepFun: Step 3.7 Flash is estimated to save $38.45/month ($461/year).

Based on 50M tokens (30% gen)

Key Selection Recommendations

1
StepFun: Step 3.7 Flash:Pick StepFun: Step 3.7 Flash when you are optimizing output cost.

Recommended Workload Routing

Repo / coding agentsinsufficient data

Need SWE-bench on both sides.

High-volume chatStepFun: Step 3.7 Flash

Lower output list price ($1.15/1M).

Voice / low-latency UIinsufficient data

Need TTFT on both sides.

Long-document RAGMoonshotAI: Kimi K2.5

Larger window (262k).

Screenshots / visionMoonshotAI: Kimi K2.5

MoonshotAI: Kimi K2.5 is the side marked multimodal in the catalog.

Target Workload	Recommended Pick	Evaluation Rationale
Repo / coding agents	insufficient data	Need SWE-bench on both sides.
High-volume chat	StepFun: Step 3.7 Flash	Lower output list price ($1.15/1M).
Voice / low-latency UI	insufficient data	Need TTFT on both sides.
Long-document RAG	MoonshotAI: Kimi K2.5	Larger window (262k).
Screenshots / vision	MoonshotAI: Kimi K2.5	MoonshotAI: Kimi K2.5 is the side marked multimodal in the catalog.

Community Sentiment

Cast Your Matchup Vote

0 total votes

Community Discussions (0)

No comments posted on this matchup yet. Be the first to share an evaluation note!

Share card

The real PNG is generated at /compare/kimi-k2-5-vs-step-3-7-flash/opengraph-image for crawlers.

PNG

CompareLLM Matrix

Verified head-to-head card

LIVE

Moonshot

MoonshotAI: Kimi K2.5

Elo 1,515

LiveBench67.2%

SWE-bench71.3%

Speed72 tok/s

Output cost$2.85/1M

stepfun

StepFun: Step 3.7 Flash

Elo n/a

LiveBench—

SWE-bench—

Speed—

Output cost$1.15/1M

Frequently Asked Questions

Which is better overall, MoonshotAI: Kimi K2.5 or StepFun: Step 3.7 Flash?: We do not have preference Elo for both models, so we do not declare an overall winner. Compare the metrics that exist.
Which is better at coding, MoonshotAI: Kimi K2.5 or StepFun: Step 3.7 Flash?: SWE-bench is missing for at least one model, so we do not rank coding from a single number.
Which is cheaper to run in production?: StepFun: Step 3.7 Flash output tokens are $1.15/1M versus $2.85/1M. Input prices and retry rates still move the real bill.
Which is faster, MoonshotAI: Kimi K2.5 or StepFun: Step 3.7 Flash?: We do not have TTFT for both models.
Which has the larger context window?: MoonshotAI: Kimi K2.5 accepts 262k tokens versus 262k.
Can I self-host either model?: Neither model is marked open-weights here. You are comparing hosted APIs.
Are these scores from CompareLLM’s own evals?: No. V1 aggregates public leaderboards (and optional first-party latency pings). Each cell names its source and date. Read /methodology.
What does preference Elo mean on this page?: Preference Elo is a crowd vote from LMArena / Arena. People see two hidden answers and pick the one they like more. The model that wins more often gets a higher Elo. That means people preferred it — not that it passed a school test. It is not SWE-bench, not accuracy, and not a number we invent. MoonshotAI: Kimi K2.5 and StepFun: Step 3.7 Flash Elo cells are dated snapshots from the named source — we do not compute Arena ourselves.

Similar Head-to-Head Comparisons

All MoonshotAI: Kimi K2.5 matchups →|All StepFun: Step 3.7 Flash matchups →

MoonshotAI: Kimi K2.5 vs MoonshotAI: Kimi K2 0711 (previous kimi)MoonshotAI: Kimi K2.5 vs Kimi K3 (next kimi)Qwen 3 Max vs MoonshotAI: Kimi K2.5 Claude Sonnet 4.5 vs MoonshotAI: Kimi K2.5 Gemini 3.6 Flash vs MoonshotAI: Kimi K2.5 Kimi Chat 1.5 vs MoonshotAI: Kimi K2.5

MoonshotAI: Kimi K2.5 vs StepFun: Step 3.7 Flash benchmark

Interactive Comparison Diagram

Matchup Breakdown

Tug-of-War Matchup Matrix

Category wins across intelligence, speed, and pricing efficiency.

MoonshotAI: Kimi K2.5 (7)vsStepFun: Step 3.7 Flash (2)

MoonshotAI: Kimi K2.5: 7W (78%)Overall: MoonshotAI: Kimi K2.5StepFun: Step 3.7 Flash: 2W (22%)

← MoonshotAI: Kimi K2.51 tiesStepFun: Step 3.7 Flash →

MoonshotAI: Kimi K2.5 (5/5)

Intelligence & Reasoning

Preference Elo, Coding proficiency, SWE-bench & LiveBench accuracy

Preference Elo

1,515vs—

Coding Elo

1,520vs—

SWE-bench

71.3%vs—

LiveBench

67.2%vs—

GPQA Diamond

87.6%vs—

MoonshotAI: Kimi K2.5: 5WStepFun: Step 3.7 Flash: 0W

MoonshotAI: Kimi K2.5 (2/2)

Speed & Latency

Generation throughput and time to first token responsiveness

Output speed

72 tok/svs—

Time to first token

270 msvs—

MoonshotAI: Kimi K2.5: 2WStepFun: Step 3.7 Flash: 0W

StepFun: Step 3.7 Flash (2/3)

Pricing & Capacity

Cost per million tokens and max context window length

Output price

$2.85/1Mvs$1.15/1M

Input price

$0.57/1Mvs$0.2/1M

Context window

262kvs262k

MoonshotAI: Kimi K2.5: 0WStepFun: Step 3.7 Flash: 2W

Recommended Workload Routing

Repo / coding agentsinsufficient data

Need SWE-bench on both sides.

High-volume chatStepFun: Step 3.7 Flash

Lower output list price ($1.15/1M).

Voice / low-latency UIinsufficient data

Need TTFT on both sides.

Long-document RAGMoonshotAI: Kimi K2.5

Larger window (262k).

Screenshots / visionMoonshotAI: Kimi K2.5

MoonshotAI: Kimi K2.5 is the side marked multimodal in the catalog.

Target Workload	Recommended Pick	Evaluation Rationale
Repo / coding agents	insufficient data	Need SWE-bench on both sides.
High-volume chat	StepFun: Step 3.7 Flash	Lower output list price ($1.15/1M).
Voice / low-latency UI	insufficient data	Need TTFT on both sides.
Long-document RAG	MoonshotAI: Kimi K2.5	Larger window (262k).
Screenshots / vision	MoonshotAI: Kimi K2.5	MoonshotAI: Kimi K2.5 is the side marked multimodal in the catalog.

Frequently Asked Questions

Which is better overall, MoonshotAI: Kimi K2.5 or StepFun: Step 3.7 Flash?

We do not have preference Elo for both models, so we do not declare an overall winner. Compare the metrics that exist.

Which is better at coding, MoonshotAI: Kimi K2.5 or StepFun: Step 3.7 Flash?

SWE-bench is missing for at least one model, so we do not rank coding from a single number.

Which is cheaper to run in production?

StepFun: Step 3.7 Flash output tokens are $1.15/1M versus $2.85/1M. Input prices and retry rates still move the real bill.

Which is faster, MoonshotAI: Kimi K2.5 or StepFun: Step 3.7 Flash?

We do not have TTFT for both models.

Which has the larger context window?

MoonshotAI: Kimi K2.5 accepts 262k tokens versus 262k.

Can I self-host either model?

Neither model is marked open-weights here. You are comparing hosted APIs.

Are these scores from CompareLLM’s own evals?

No. V1 aggregates public leaderboards (and optional first-party latency pings). Each cell names its source and date. Read /methodology.

What does preference Elo mean on this page?

Preference Elo is a crowd vote from LMArena / Arena. People see two hidden answers and pick the one they like more. The model that wins more often gets a higher Elo. That means people preferred it — not that it passed a school test. It is not SWE-bench, not accuracy, and not a number we invent. MoonshotAI: Kimi K2.5 and StepFun: Step 3.7 Flash Elo cells are dated snapshots from the named source — we do not compute Arena ourselves.