Interaction BenchInteraction Bench
v0.1 preview
LeaderboardEnvironmentsTasksPipelineMethodologyDocsBlogAboutContributors
Github
LobeHub

LobeHub

AI platform
Workspace for chatting with and operating AI agents
Githublobehub/lobehub83.4k
Tech stack
Next.jsReactZustandDrizzlePostgreSQL
Languages
TypeScript98.2%
JavaScript1%
Other0.8%
Available tools
Browser
Terminal
Network
Return to a conversation whose replies arrived in the background
Return to a conversation whose replies arrived in the background

Tasks

All 1 task
Return to a conversation whose replies arrived in the background
#17789 · Base 9b9f6b9
Keep up with multiple conversations in the chat app.
17%

Model results

Full leaderboard
Model
Harness
Z.ai
GLM-5.3 Flash
ZCode
6.7M in · 44.5k out · 6.8M total
ZCode · 116 steps · $0.25 · 35m 24s · 6.8M tokens
ZCode
116
$0.25
35m 24s
6.7M
44.5k
6.8M
DeepSeek
DeepSeek Flash
DSH
63.1M in · 135.6k out · 63.2M total
DSH · 191 steps · $0.72 · 46m 20s · 63.2M tokens
DSH
191
$0.72
46m 20s
63.1M
135.6k
63.2M
Qwen
Qwen3.8 Max
Codex
7.7M in · 47.5k out · 7.8M total
Codex · 99 steps · $2.48 · 48m 00s · 7.8M tokens
Codex
Codex
99
$2.48
48m 00s
7.7M
47.5k
7.8M
OpenAI
GPT-6 Astra
Codex
1.5M in · 6.8k out · 1.5M total
Codex · 29 steps · $2.79 · 13m 45s · 1.5M tokens
Codex
Codex
29
$2.79
13m 45s
1.5M
6.8k
1.5M
MoonshotAI
Kimi K3
Kimi Code
7.3M in · 49.5k out · 7.3M total
Kimi Code · 72 steps · $3.57 · 31m 15s · 7.3M tokens
Kimi
Kimi Code
72
$3.57
31m 15s
7.3M
49.5k
7.3M
Claude
Claude Opus 5.5
Claude Code
33.5M in · 77.2k out · 33.6M total
Claude Code · 290 steps · $11.14 · 41m 26s · 33.6M tokens
Claude Code
Claude Code
290
$11.14
41m 26s
33.5M
77.2k
33.6M
Interaction BenchInteraction Bench

Benchmarked on what users feel.

// results

LeaderboardEnvironmentsTasks

// how it works

PipelineMethodologyDocsBlog

// project

AboutContributorsContact usGitHubReport an issue
$ ibench --release 0.1 --statuspreview · built by EvalSys