Multi-turn model evaluation

Multi-Turn Benchmark Leaderboard

This multi-turn benchmark uses three tasks: bug-fix-02, refactor-02, and migration-02, each split across two staged turns for six total stages. Requirements are withheld until each stage is reached. Ranking order is: scored passes desc, strict passes desc, soft passes desc, completed stages desc, repairs used asc, total output tokens asc, total wall time asc, total cost asc, then stage-1 initial passes desc.

LegendPStrict passSSoft passF:*Failed case
Showing 39 of 39
Rank
1
Model
kimi-k2.7-codemoonshot
Result
3/3 Strict / 3/3 ScoredSoft 0/3 · Hard 0/3
Stages
6/6
Init
3/3
Repairs
0/3
Output
46685
Cost
$0.34
Wall
811.5
Cases
Bug FixPRefactorPMigrationP
Rank
2
Model
gpt-5.6-terraopenai
Result
3/3 Strict / 3/3 ScoredSoft 0/3 · Hard 0/3
Stages
6/6
Init
2/3
Repairs
1/3
Output
11151
Cost
$0.60
Wall
94.4
Cases
Bug FixPRefactorPMigrationP
Rank
3
Model
gemini-3.5-flashgoogle
Result
3/3 Strict / 3/3 ScoredSoft 0/3 · Hard 0/3
Stages
6/6
Init
2/3
Repairs
1/3
Output
13208
Cost
$0.55
Wall
333.1
Cases
Bug FixPRefactorPMigrationP
Rank
4
Model
gpt-5.4openai
Result
3/3 Strict / 3/3 ScoredSoft 0/3 · Hard 0/3
Stages
6/6
Init
2/3
Repairs
1/3
Output
13467
Cost
$0.70
Wall
144.8
Cases
Bug FixPRefactorPMigrationP
Rank
5
Model
qwen3.5-plusqwen
Result
3/3 Strict / 3/3 ScoredSoft 0/3 · Hard 0/3
Stages
6/6
Init
2/3
Repairs
1/3
Output
16719
Cost
$0.12
Wall
186.5
Cases
Bug FixPRefactorPMigrationP
Rank
6
Model
claude-opus-5anthropic
Result
3/3 Strict / 3/3 ScoredSoft 0/3 · Hard 0/3
Stages
6/6
Init
2/3
Repairs
1/3
Output
21267
Cost
$1.82
Wall
203.9
Cases
Bug FixPRefactorPMigrationP
Rank
7
Model
gpt-5.3-codexopenai
Result
3/3 Strict / 3/3 ScoredSoft 0/3 · Hard 0/3
Stages
6/6
Init
2/3
Repairs
1/3
Output
23459
Cost
$0.63
Wall
367.0
Cases
Bug FixPRefactorPMigrationP
Rank
8
Model
GPT-5.5openai
Result
3/3 Strict / 3/3 ScoredSoft 0/3 · Hard 0/3
Stages
6/6
Init
2/3
Repairs
1/3
Output
26611
Cost
$1.65
Wall
328.2
Cases
Bug FixPRefactorPMigrationP
Rank
9
Model
deepseek-v4-prodeepseek
Result
3/3 Strict / 3/3 ScoredSoft 0/3 · Hard 0/3
Stages
6/6
Init
2/3
Repairs
1/3
Output
30324
Cost
$0.11
Wall
344.0
Cases
Bug FixPRefactorPMigrationP
Rank
10
Model
kimi-k2.5moonshot
Result
3/3 Strict / 3/3 ScoredSoft 0/3 · Hard 0/3
Stages
6/6
Init
2/3
Repairs
1/3
Output
40544
Cost
$0.23
Wall
935.6
Cases
Bug FixPRefactorPMigrationP
Rank
11
Model
kimi-k2.7-code-highspeedmoonshot
Result
3/3 Strict / 3/3 ScoredSoft 0/3 · Hard 0/3
Stages
6/6
Init
2/3
Repairs
1/3
Output
41530
Cost
$0.69
Wall
264.4
Cases
Bug FixPRefactorPMigrationP
Rank
12
Model
qwen3.6-plus-2026-04-02qwen
Result
3/3 Strict / 3/3 ScoredSoft 0/3 · Hard 0/3
Stages
6/6
Init
2/3
Repairs
1/3
Output
53757
Cost
$0.27
Wall
996.8
Cases
Bug FixPRefactorPMigrationP
Rank
13
Model
qwen3.7-maxqwen
Result
3/3 Strict / 3/3 ScoredSoft 0/3 · Hard 0/3
Stages
6/6
Init
2/3
Repairs
1/3
Output
58931
Cost
$0.82
Wall
1090.0
Cases
Bug FixPRefactorPMigrationP
Rank
14
Model
gemini-3.1-pro-previewgoogle
Result
3/3 Strict / 3/3 ScoredSoft 0/3 · Hard 0/3
Stages
6/6
Init
2/3
Repairs
2/3
Output
13083
Cost
$0.71
Wall
1253.1
Cases
Bug FixPRefactorPMigrationP
Rank
15
Model
grok-4.3grok
Result
3/3 Strict / 3/3 ScoredSoft 0/3 · Hard 0/3
Stages
6/6
Init
1/3
Repairs
2/3
Output
24055
Cost
$0.30
Wall
305.0
Cases
Bug FixPRefactorPMigrationP
Rank
16
Model
glm-5.1zai
Result
3/3 Strict / 3/3 ScoredSoft 0/3 · Hard 0/3
Stages
6/6
Init
1/3
Repairs
2/3
Output
53999
Cost
$0.59
Wall
1083.1
Cases
Bug FixPRefactorPMigrationP
Rank
17
Model
glm-5.2zai
Result
3/3 Strict / 3/3 ScoredSoft 0/3 · Hard 0/3
Stages
6/6
Init
2/3
Repairs
2/3
Output
85545
Cost
$0.68
Wall
1022.5
Cases
Bug FixPRefactorPMigrationP
Rank
18
Model
claude-opus-4-7anthropic
Result
3/3 Strict / 3/3 ScoredSoft 0/3 · Hard 0/3
Stages
6/6
Init
1/3
Repairs
3/3
Output
15499
Cost
$2.02
Wall
145.5
Cases
Bug FixPRefactorPMigrationP
Rank
19
Model
glm-5zai
Result
2/3 Strict / 3/3 ScoredSoft 1/3 · Hard 0/3
Stages
5/6
Init
1/3
Repairs
2/3
Output
40943
Cost
$0.40
Wall
785.8
Cases
Bug FixPRefactorSMigrationP
Rank
20
Model
claude-opus-4-8anthropic
Result
2/3 Strict / 2/3 ScoredSoft 0/3 · Hard 1/3
Stages
5/6
Init
1/3
Repairs
2/3
Output
16099
Cost
$2.21
Wall
148.8
Cases
Bug FixPRefactorF:FORMATMigrationP
Rank
21
Model
gpt-5.6-solopenai
Result
2/3 Strict / 2/3 ScoredSoft 0/3 · Hard 1/3
Stages
4/6
Init
2/3
Repairs
0/3
Output
10079
Cost
$0.96
Wall
129.2
Cases
Bug FixPRefactorF:STRUCTURE_FAILMigrationP
Rank
22
Model
gpt-5.6-lunaopenai
Result
2/3 Strict / 2/3 ScoredSoft 0/3 · Hard 1/3
Stages
4/6
Init
2/3
Repairs
0/3
Output
10722
Cost
$0.19
Wall
65.9
Cases
Bug FixPRefactorF:STRUCTURE_FAILMigrationP
Rank
23
Model
grok-4.5grok
Result
2/3 Strict / 2/3 ScoredSoft 0/3 · Hard 1/3
Stages
4/6
Init
2/3
Repairs
0/3
Output
11891
Cost
$0.23
Wall
76.7
Cases
Bug FixPRefactorF:STRUCTURE_FAILMigrationP
Rank
24
Model
claude-fable-5anthropic
Result
2/3 Strict / 2/3 ScoredSoft 0/3 · Hard 1/3
Stages
4/6
Init
2/3
Repairs
0/3
Output
12334
Cost
$2.38
Wall
136.2
Cases
Bug FixPRefactorF:STRUCTURE_FAILMigrationP
Rank
25
Model
qwen3.7-plusqwen
Result
2/3 Strict / 2/3 ScoredSoft 0/3 · Hard 1/3
Stages
4/6
Init
2/3
Repairs
0/3
Output
28872
Cost
$0.17
Wall
556.5
Cases
Bug FixPRefactorF:STRUCTURE_FAILMigrationP
Rank
26
Model
qwen3.6-max-previewqwen
Result
2/3 Strict / 2/3 ScoredSoft 0/3 · Hard 1/3
Stages
4/6
Init
2/3
Repairs
0/3
Output
36347
Cost
$0.54
Wall
893.9
Cases
Bug FixPRefactorF:STRUCTURE_FAILMigrationP
Rank
27
Model
kimi-k3moonshot
Result
2/3 Strict / 2/3 ScoredSoft 0/3 · Hard 1/3
Stages
4/6
Init
2/3
Repairs
0/3
Output
42088
Cost
$1.05
Wall
1431.2
Cases
Bug FixPRefactorF:STRUCTURE_FAILMigrationP
Rank
28
Model
kimi-k2.6moonshot
Result
2/3 Strict / 2/3 ScoredSoft 0/3 · Hard 1/3
Stages
4/6
Init
2/3
Repairs
0/3
Output
69038
Cost
$0.40
Wall
1942.9
Cases
Bug FixPRefactorF:STRUCTURE_FAILMigrationP
Rank
29
Model
deepseek-v4-flashdeepseek
Result
2/3 Strict / 2/3 ScoredSoft 0/3 · Hard 1/3
Stages
4/6
Init
1/3
Repairs
1/3
Output
16890
Cost
$0.03
Wall
162.3
Cases
Bug FixPRefactorF:STRUCTURE_FAILMigrationP
Rank
30
Model
glm-5-turbozai
Result
2/3 Strict / 2/3 ScoredSoft 0/3 · Hard 1/3
Stages
4/6
Init
1/3
Repairs
1/3
Output
17196
Cost
$0.27
Wall
289.7
Cases
Bug FixPRefactorF:STRUCTURE_FAILMigrationP
Rank
31
Model
claude-sonnet-5anthropic
Result
2/3 Strict / 2/3 ScoredSoft 0/3 · Hard 1/3
Stages
4/6
Init
1/3
Repairs
1/3
Output
23304
Cost
$1.04
Wall
175.4
Cases
Bug FixPRefactorF:STRUCTURE_FAILMigrationP
Rank
32
Model
grok-4.20-0309-reasoninggrok
Result
2/3 Strict / 2/3 ScoredSoft 0/3 · Hard 1/3
Stages
4/6
Init
1/3
Repairs
1/3
Output
34710
Cost
$0.52
Wall
170.5
Cases
Bug FixPRefactorF:STRUCTURE_FAILMigrationP
Rank
33
Model
grok-build-0.1grok
Result
2/3 Strict / 2/3 ScoredSoft 0/3 · Hard 1/3
Stages
4/6
Init
1/3
Repairs
1/3
Output
47900
Cost
$0.26
Wall
247.2
Cases
Bug FixPRefactorF:STRUCTURE_FAILMigrationP
Rank
34
Model
gemini-3.1-flash-litegoogle
Result
2/3 Strict / 2/3 ScoredSoft 0/3 · Hard 1/3
Stages
4/6
Init
0/3
Repairs
2/3
Output
12158
Cost
$0.08
Wall
41.8
Cases
Bug FixPRefactorF:STRUCTURE_FAILMigrationP
Rank
35
Model
claude-opus-4-6anthropic
Result
2/3 Strict / 2/3 ScoredSoft 0/3 · Hard 1/3
Stages
4/6
Init
1/3
Repairs
2/3
Output
17122
Cost
$1.47
Wall
167.3
Cases
Bug FixPRefactorF:STRUCTURE_FAILMigrationP
Rank
36
Model
deepseek-reasonerdeepseek
Result
1/3 Strict / 1/3 ScoredSoft 0/3 · Hard 2/3
Stages
3/6
Init
1/3
Repairs
2/3
Output
21980
Cost
$0.02
Wall
676.3
Cases
Bug FixPRefactorF:STRUCTURE_FAILMigrationF:FORMAT
Rank
37
Model
deepseek-chatdeepseek
Result
1/3 Strict / 1/3 ScoredSoft 0/3 · Hard 2/3
Stages
3/6
Init
0/3
Repairs
3/3
Output
19806
Cost
$0.11
Wall
537.1
Cases
Bug FixF:TEST_FAILRefactorF:TEST_FAILMigrationP
Rank
38
Model
grok-4.20-beta-latest-non-reasoninggrok
Result
1/3 Strict / 1/3 ScoredSoft 0/3 · Hard 2/3
Stages
2/6
Init
1/3
Repairs
1/3
Output
9499
Cost
$0.38
Wall
48.6
Cases
Bug FixF:TEST_FAILRefactorF:STRUCTURE_FAILMigrationP
Rank
39
Model
MiniMax-M2.7minimax
Result
0/3 Strict / 1/3 ScoredSoft 1/3 · Hard 2/3
Stages
0/6
Init
0/3
Repairs
3/3
Output
43042
Cost
$0.09
Wall
988.2
Cases
Bug FixF:FORMATRefactorF:FORMATMigrationS

Numeric columns are sortable. Currency is rounded to two decimals for readability.