Gemini 3.8 Flash achieves 73.7% on DeepSWE.
It pushes the frontier with an impressive 8.2% increase over Gemini 3.7 Flash, at the same cost but with more steps and output tokens.
It pushes the frontier with an impressive 8.2% increase over Gemini 3.7 Flash, at the same cost but with more steps and output tokens.
40
27
478
71,422
Gemini 3.7 Flash debuts at 65.5% on DeepSWE.
It delivers substantial improvements over 3.6 Flash, scoring +18.8% higher while costing less than half as much per task.
It delivers substantial improvements over 3.6 Flash, scoring +18.8% higher while costing less than half as much per task.
47
53
798
225,629
View the full benchmark results on DeepSWE → deepswe.datacurve.ai
0
1
17
5,796
Claude Opus 5's results are now on DeepSWE
With a score of 74%, it's the best long-horizon coding model we've seen.
With a score of 74%, it's the best long-horizon coding model we've seen.
104
56
1,380
246,121
Gemini 3.6 Flash is now on DeepSWE at 49%, roughly on par with Opus 4.8 at medium reasoning.
51
51
1,314
249,726
Gemini 3.6 Flash is significantly more efficient while scoring higher.
It uses 65% fewer output tokens per task and costs 52% less per task compared to its predecessor.
It uses 65% fewer output tokens per task and costs 52% less per task compared to its predecessor.
2
2
94
9,692
Kimi K3 debuts at #3 on DeepSWE.
It's the first open-weights model that delivers frontier-level performance, achieving results similar to Claude Fable and GPT-5.6 Sol.
It's the first open-weights model that delivers frontier-level performance, achieving results similar to Claude Fable and GPT-5.6 Sol.
106
339
3,651
670,458
View the full leaderboard & benchmark. → deepswe.datacurve.ai
2
1
38
12,432
GPT-5.6 tops the DeepSWE leaderboard at 73%.
Sol, Terra, and Luna results are now available.
Sol, Terra, and Luna results are now available.
71
168
2,026
348,087
View the full family of GPT-5.6 models on our benchmark → deepswe.datacurve.ai
4
6
79
26,803
Datacurve is at @aiDotEngineer!
Come by our booth in the expo hall to learn more about DeepSWE (and snag some merch).
Plus, join us on Thursday @ 10:30am for our talk.
Come by our booth in the expo hall to learn more about DeepSWE (and snag some merch).
Plus, join us on Thursday @ 10:30am for our talk.
1
3
42
7,655
GLM 5.2 is now on DeepSWE as the top open-source model on our leaderboard.
With a pass@1 score of 44% at max effort, GLM 5.2 is indisputable #1 open-source model besting Kimi K2.7 Code by 17%.
With a pass@1 score of 44% at max effort, GLM 5.2 is indisputable #1 open-source model besting Kimi K2.7 Code by 17%.
103
237
2,779
586,956
Claude Fable 5 debuts at #1 on DeepSWE.
It outscores the previous best by 3% and sets a new state-of-the-art on our long-horizon coding benchmark.
It outscores the previous best by 3% and sets a new state-of-the-art on our long-horizon coding benchmark.
108
84
1,617
476,090
Fable 5 scores 70% pass@1 and tracks GPT-5.5 on cost-performance at the default high effort.
Kimi K2.7 also joins the leaderboard with a score of 31%.
Kimi K2.7 also joins the leaderboard with a score of 31%.
9
8
136
26,200
See the full updated leaderboard here: deepswe.datacurve.ai
5
2
56
14,225