Board · task version 1

hard-review

Each board ranks harness × model combinations on one suite by pass rate with a 95% Wilson interval; rows are not comparable across suites or task versions.

9 tasks
  • hard-review-clean-long
  • hard-review-cross-file
  • hard-review-cross-file-v2
  • hard-review-decoy-decision-replies
  • hard-review-decoy-job-leases
  • hard-review-multi-defect-outbox
  • hard-review-multi-defect-outbox-v2
  • hard-review-multi-defect-tool-calls
  • hard-review-multi-defect-tool-calls-v2
hard-review · task version 1
#Liverymodel · harness · configpassed / n95% Wilsonmeantaskstokens / passcost / passexcluded
1together/deepseek-v4.1-flashpi · config da41c66e13 / 13 77%–100%0.936244k$0.09 list0
2together/deepseek-v4.1-flashopencode · config 357da8919 / 10 60%–98%0.859not measured (3/10)unknown5
3zai/glm-5.2codex · config 3437a9e616 / 18 67%–97%0.878490k$0.48 list1
4zai/glm-5.2claude-code · config ecb67e4e8 / 9 57%–98%0.849not measured (5/9)unknown11
5google/gemini-3.8-flash @highantigravity · config 034ed3a121 / 24 69%–96%0.9481.77M$0.59 list0
6google/gemini-4-argon @highantigravity · config 034ed3a17 / 8 53%–98%0.898272k$0.41 list0
7fireworks/kimi-k3opencode · config 357da8917 / 8 53%–98%0.898231k$0.38 list0
8zai/glm-5.2opencode · config 357da8917 / 8 53%–98%0.858294k$0.26 list0
9fireworks/kimi-k3pi · config da41c66e7 / 8 53%–98%0.848159k$0.32 list0
10google/gemini-3.8-flashpi · config da41c66e7 / 8 53%–98%0.9281.08M$0.42 list0
11google/gemini-3.8-flash @highpi · config da41c66e7 / 8 53%–98%0.9481.89M$0.64 list0
12moonshot/kimi-k3pi · config da41c66e7 / 8 53%–98%0.828147k$0.28 list0
13openai/gpt-6.1-solpi · config da41c66e7 / 8 53%–98%0.96877k$0.07 list0
14zai/glm-5.2pi · config da41c66e7 / 8 53%–98%0.848305k$0.28 list0
15together/deepseek-v4.1-flashclaude-code · config ecb67e4e15 / 18 61%–94%0.898816k$0.29 list0
16baseten/glm-5.3-flashopencode · config 357da89115 / 18 61%–94%0.898251k$0.03 list1
17baseten/glm-5.3-flashpi · config da41c66e14 / 18 55%–91%0.868203k$0.03 list0
18fireworks/qwen3.8-maxpi · config da41c66e14 / 18 55%–91%0.898377k$0.30 list0
19openai/gpt-6.1-sol @mediumcodex · config b239640338 / 50 63%–86%0.945185k$0.09 list0
20baseten/glm-5.3-flashclaude-code · config ecb67e4e6 / 8 41%–93%0.8781.27M$0.09 list0
21fireworks/kimi-k3claude-code · config ecb67e4e6 / 8 41%–93%0.858327k$0.24 list0
22openai/gpt-6.1-solclaude-code · config ecb67e4e6 / 8 41%–93%0.958390k$0.21 list0
23google/gemini-3.8-flashcodex · config 3437a9e66 / 8 41%–93%0.9181.28M$0.49 list0
24anthropic/claude-sonnet-5-5opencode · config 357da8916 / 8 41%–93%0.818164k$0.14 list1
25fireworks/qwen3.8-maxopencode · config 357da8916 / 8 41%–93%0.948396k$0.34 list0
26moonshot/kimi-k3opencode · config 357da8916 / 8 41%–93%0.808244k$0.40 list0
27fireworks/qwen3.8-maxclaude-code · config ecb67e4e16 / 22 52%–87%0.909not measured (13/22)unknown11
28openai/gpt-6.1-solcodex · config b239640320 / 29 51%–83%0.929not measured (23/29)unknown0
29google/gemini-3.8-flashclaude-code · config ecb67e4e5 / 8 31%–86%0.9282.40M$0.71 list0
30moonshot/kimi-k3claude-code · config ecb67e4e5 / 8 31%–86%0.798599k$0.39 list0
31anthropic/claude-opus-5-5codex · config 3437a9e65 / 8 31%–86%0.918145k$0.65 list0
32moonshot/kimi-k3codex · config 3437a9e65 / 8 31%–86%0.768248k$0.32 list0
33google/gemini-3.8-flashopencode · config 357da8915 / 8 31%–86%0.8381.45M$0.53 list0
34anthropic/claude-sonnet-5-5pi · config da41c66e5 / 8 31%–86%0.82898k$0.24 list0
35fireworks/kimi-k3codex · config 3437a9e611 / 18 39%–80%0.798210k$0.29 list0
36openai/gpt-6.1-sol @xhighcodex · config b239640314 / 25 37%–73%0.905336k$0.20 list0
37openai/gpt-6.1-solopencode · config 357da8919 / 18 29%–71%0.898437k$0.27 list0
38anthropic/claude-sonnet-5-5codex · config 3437a9e64 / 8 22%–78%0.758168k$0.37 list0
39fireworks/qwen3.8-maxcodex · config 3437a9e64 / 8 22%–78%0.878594k$0.61 list8
40anthropic/claude-opus-5-5pi · config da41c66e4 / 8 22%–78%0.898117k$0.58 list0
41openai/gpt-6.1-sol @highcodex · config b239640317 / 35 33%–64%0.878320k$0.18 list7
42google/gemini-3.1-pro @highantigravity · config 034ed3a14 / 9 19%–73%0.709not measured (3/9)unknown0
43anthropic/claude-opus-5-5 @xhighclaude-code · config c32a950b8 / 20 22%–61%0.8554.53M$2.78 list3
44anthropic/claude-opus-5-5claude-code · config c32a950b11 / 28 24%–58%0.8583.00M$1.20 list0
45anthropic/claude-sonnet-5-5claude-code · config c32a950b3 / 8 14%–69%0.8082.53M$0.55 list0
46baseten/glm-5.3-flashcodex · config 3437a9e63 / 8 14%–69%0.768485k$0.07 list0
47anthropic/claude-opus-5-5opencode · config 357da8913 / 8 14%–69%0.848262k$0.52 list0
48together/deepseek-v4.1-flashcodex · config 3437a9e60 / 3 0%–56%0.703no passesunknown0
list = published-rate · invoice = provider-invoice-reconciled