Skip to content
ben ebsworth
projects
blog
lab
vectors
hiking
about
Search
⌘K
Home
/
Lab
ben ebsworth
projects
blog
lab
vectors
hiking
about
Search
⌘K
Home
/
Lab
/
Harness Eval
/
Models
/
qwen3.8-max-preview
qwen
qwen3.8-max-preview
Pass rate: 70%
14/20 passed
Avg: 2m 55s
Tokens: 2.3M
Cost: $0.72
Per-task results
Task
Suite
Status
Duration
Tokens
Cost
Turns
Patch
Find and fix an event listener leak
hard-targeting
pass
1m 53s
88.2k
$0.029
—
yes
Rename a public function across the codebase
hard-targeting
done
1m 15s
76.3k
$0.025
—
yes
Fix concurrent counter race condition
hard-targeting
done
2m 47s
67.6k
$0.022
—
yes
Restore linear performance in a hot loop
hard-targeting
pass
5m 17s
66.0k
$0.021
—
yes
Add an options parameter to format() without breaking callers
hard-targeting
done
2m 39s
216.0k
$0.070
—
yes
Read SPEC.md and implement the right behavior
hard-targeting
pass
1m 21s
79.4k
$0.026
—
yes
Two parsers disagree on the same config file
hard-targeting
done
3m 10s
125.8k
$0.041
—
yes
Migrate from fs.readFile(callback) to fs.promises.readFile()
hard-targeting
done
7m 43s
215.9k
$0.070
—
yes
Fix implicit string→number coercion in comparison
hard-targeting
pass
3m 21s
112.4k
$0.037
—
yes
A test passes for the wrong reason — find the real fix
hard-targeting
done
5m 11s
102.2k
$0.033
—
yes
Find and fix an event listener leak
hard-targeting
pass
1m 34s
77.9k
$0.025
—
yes
Rename a public function across the codebase
hard-targeting
pass
1m 37s
140.5k
$0.041
—
yes
Fix concurrent counter race condition
hard-targeting
pass
2m 7s
67.7k
$0.023
—
yes
Restore linear performance in a hot loop
hard-targeting
pass
3m 20s
92.8k
$0.027
—
yes
Add an options parameter to format() without breaking callers
hard-targeting
pass
2m 1s
191.0k
$0.055
—
yes
Read SPEC.md and implement the right behavior
hard-targeting
pass
2m 11s
82.5k
$0.028
—
yes
Two parsers disagree on the same config file
hard-targeting
pass
1m 20s
108.6k
$0.032
—
yes
Migrate from fs.readFile(callback) to fs.promises.readFile()
hard-targeting
pass
3m 47s
154.1k
$0.052
—
yes
Fix implicit string→number coercion in comparison
hard-targeting
pass
4m 6s
122.8k
$0.037
—
yes
A test passes for the wrong reason — find the real fix
hard-targeting
pass
1m 44s
83.5k
$0.027
—
yes
All models