Kimi K3 vs Muse Spark 1.3

At a glance

• Capability index

  • Kimi K3157.4
  • Muse Spark 1.3156.8

  Kimi K3 is 0.7 points higher.

• Price per 1M tokens

  • Kimi K3$6.00
  • Muse Spark 1.3$2.00

  Muse Spark 1.3 is 3 times cheaper.

• Context window

  • Kimi K31M tokens
  • Muse Spark 1.31M tokens

  They take in about the same.

• Released

  • Kimi K3Jul 16, 2026
  • Muse Spark 1.3Sep 2, 2026

  Muse Spark 1.3 is 48 days newer.

Where each scores higher

On the 19 tests both have taken, each on its own scale. A gap of a point or two can sit within a test’s margin of error.

Level on Arena hard prompts.

4 more tests have results for only one of them; the table below lists every result.

Everything side by side

Comparison of Kimi K3 vs Muse Spark 1.3

Measure	Kimi K3Moonshot AI	Muse Spark 1.3Meta AI
Model
Maker	Moonshot AI	Meta AI
Sold here by	Moonshot AI	DevPass (LLM Gateway)
API model ID	kimi-k3	muse-spark-1.3
Released	Jul 16, 2026	Sep 2, 2026
Knowledge cutoff	Not reported	Not reported
Weights	Open: downloadable	Closed: hosted access only
Developer’s country	China	United States
Hosts selling it	84including Moonshot AI directly	11
Training compute	2.0 × 10²⁵ FLOPEpoch AI estimate	Not published
Test results
Capability index	157.4 (best of these models)#15 of 274	156.8#19 of 274
LiveBench overall	79.2%#15 of 66	81.6% (best of these models)#7 of 66
Arena text	1488#16 of 413	1494 (best of these models)#9 of 413
LiveBench coding	81.5% (best of these models)#15 of 66	81.1%#17 of 66
Arena WebDev	1658 (best of these models)#13 of 138	1657#14 of 138
Arena coding	1541 (best of these models)#9 of 408	1539#11 of 408
LiveBench mathematics	84.4%#53 of 66	96.0% (best of these models)#12 of 66
Mock AIME 2024–2025	97.2%#31 of 297	99.2% (best of these models)#15 of 297
FrontierMath Tiers 1–3	72.2%#22 of 114	74.4% (best of these models)#19 of 114
FrontierMath Tier 4	39.0%#29 of 70	46.3% (best of these models)#25 of 70
Arena math	1501#16 of 396	1509 (best of these models)#9 of 396
SimpleQA Verified	50.6%#24 of 86	Not tested
LiveBench agentic coding	62.2%#10 of 66	64.1% (best of these models)#9 of 66
DeepSWE	68.5%#19 of 69	Not tested
CursorBench	Not tested	41.6%#22 of 62
LiveBench language	85.5% (best of these models)#11 of 66	82.8%#24 of 66
LiveBench instructions	71.4%#23 of 66	78.0% (best of these models)#4 of 66
Arena creative writing	1460 (best of these models)#26 of 411	1459#28 of 411
Arena instruction following	1488 (best of these models)#14 of 413	1486#16 of 413
LiveBench data analysis	78.7%#21 of 66	79.6% (best of these models)#12 of 66
LiveBench reasoning	90.7% (best of these models)#9 of 66	89.7%#15 of 66
GPQA Diamond	93.1%#20 of 319	Not tested
Chess puzzles	39.0% (best of these models)#31 of 227	38.0%#32 of 227
Arena hard prompts	1517#9 of 413	1517#10 of 413
Price per million tokens
Input	$3.00	$1.25 (best of these models)
Output	$15.00	$4.25 (best of these models)
Blended, 3 input to 1 output	$6.00	$2.00 (best of these models)
Cached input	$0.30	$0.15 (best of these models)
The maker’s own price	This listing	Not sold directly
Limits
Context window	1M tokens	1M tokens
Max input	1M tokens	1M tokens
Max output	1M tokens (best of these models)	131.1K tokens
Features
Reads	images, text and video	audio, images, PDFs, text and video
Produces	text	text
Reasoning	Yeseffort low, high, max; can be switched off; thinking budget can be set	Yeseffort minimal, low, medium, high, xhigh
Tool calling	Yes	Yes
Structured output	Yes, such as JSON	Yes, such as JSON
File attachments	Yes	Yes
Temperature setting	Fixed by the provider	Supported

Bold marks the better value in each row. Prices are each model’s own list price, or the middle price across hosts when the maker does not sell it, blended as three input tokens for every output token. Test results as published by LiveBench (Apache-2.0), Epoch AI (CC BY 4.0), Datacurve DeepSWE, compiled by Epoch AI (CC BY 4.0), Cursor CursorBench, compiled by Epoch AI (CC BY 4.0), Arena (CC BY 4.0); model details and prices from models.dev (MIT). Confirm pricing with the provider before use.