SWE-2 Flash
- Input
- —
- Output
- —
- Speed
- —
- Context
- —
Cognition's coding model for Devin, built on Thinking Machines' Inkling-Small (276B total / 12B active) with added mid-training and RL. Weights not public (a program to share them with trusted partners is planned). Free through 11/30 in Devin Desktop and CLI; no API price published (about 3x faster and one-fifth the price of SWE-2).
Rankings
Benchmark scores
~ italic, dashed = no published score yet, estimated from related benchmarks.
Coding
SWE-bench VerifiedEstimated
~#2 of 21 (estimated)~85.8%
No published score — estimated from related benchmarks
SWE-bench Pro (Public)Estimated
~#7 of 24 (estimated)~66.8%
- #1Claude Fable 5.293.0%
- #2Claude Opus 5.589.9%
- #6Intern-S2-397B68.5%
- ~#7SWE-2 FlashEstimated66.8%
- #7Beam 1.0 501B A23B66.1%
No published score — estimated from related benchmarks
DeepSWE v1.1Lab-reported
#30 of 4560.1%
- #1Gemini 4 Argon78.8%
- #2DeepSeek V4.1 Flash 101777.0%
- #29GLM-5.361.4%
- #30SWE-2 Flash60.1%
- #31Muse Spark Open60.0%
cognition.com · 2026-11-04
Terminal-Bench 4.0Lab-reported
#30 of 10120.0%
- #1Kimi K3.1 Pro67.2%
- #2Claude Sonnet 5.563.6%
- #29MiMo-V2.6-Flash22.7%
- #30SWE-2 Flash20.0%
- #31Gemini 3.8 Flash19.7%
cognition.com · 2026-11-04
LiveCodeBenchEstimated
~#6 of 28 (estimated)~88.4%
- #1Qwen3.8-Omni-Flash92.6%
- #2Qwen3.8 Flash-Next91.9%
- #5Nemotron 3 Ultra89.0%
- ~#6SWE-2 FlashEstimated88.4%
- #6Beam 1.0 501B A23B88.3%
No published score — estimated from related benchmarks
SciCodeEstimated
~#46 of 119 (estimated)~49.6%
- #1Claude Fable 5.267.7%
- #2Claude Opus 5.566.9%
- #45Inkling Small49.7%
- ~#46SWE-2 FlashEstimated49.6%
- #46Beam 1.0 501B A23B49.2%
No published score — estimated from related benchmarks
AA Coding Agent IndexEstimated
~#21 of 26 (estimated)~40.8
- #1Claude Opus 5.561.7
- #2Claude Opus 559.7
- #20Kimi K341.4
- ~#21SWE-2 FlashEstimated40.8
- #21Qwen3.8 Max (0902)39.2
No published score — estimated from related benchmarks