PinchBench

🦞🦀 Benchmarking AI coding agents one task at a time. Daily leaderboard updates. Open source. Built by @olearycrew, powered by @kilocode

[PinchBench @pinchbench]

The numbers are the numbers. Nemotron 3 Ultra is #1 among open-weight models on PinchBench right now. This will be a good session to catch if you want the architecture behind the score. 👀

[Kilo @kilocode]

NVIDIA's Nemotron 3 Ultra is currently the #1 Open-Weight model on @pinchbench, the OpenClaw-focused model benchmark.

We're sitting down with the @NVIDIA team in a free, live webinar tomorrow (June 10th) at 1pm ET/10am PT to get an inside look at the architecture, the tradeoffs, and how Ultra slots into multi-model agent workflows in Kilo.

Register: https://t.co/GfFQ0QYa1Y

[PinchBench @pinchbench]

@ctnzr Appreciate the shoutout! Nemotron 3 Ultra earned it – 89.9 percent average puts it at the top of the open-weights board. Exactly the kind of agentic capability PinchBench is built to measure 💪

[PinchBench @pinchbench]

The moment, timestamped: https://t.co/JHOD14SzIJ

[PinchBench @pinchbench]

When the company the whole industry runs on reaches for your benchmark to make their case, the methodology is holding up. Proud to see us on that stage!

[Kilo @kilocode]

Look closely at the NVIDIA keynote slide behind Jensen Huang.

@pinchbench is at the top of it! @NVIDIA used our benchmark to position Nemotron 3 Ultra against the frontier. Nemotron tied for the lead on Agent Productivity at 91%.

Built for the agentic era. @NVIDIAAIDev, the benchmark is doing its job.

[Kilo @kilocode]

Gemini 3.5 Flash was live in Kilo before I/O 2026 ended.

74.2% on @pinchbench in initial runs. 1M token context. Roughly 4x faster than comparable frontier models, and it beats Gemini 3.1 Pro on most coding and agentic benchmarks.

Try it now: https://t.co/OnJ9NkrEsw

[PinchBench @pinchbench]

✅ Merged on pinchbench/leaderboard: "feat: add recent runs stream page"

https://t.co/jfbiLyUgT6

[PinchBench @pinchbench]

📊 PinchBench Daily Leaderboard (Average Score)

🥇 xiaomi/mimo-v2.5 — 89.2%
🥈 xiaomi/mimo-v2.5-pro — 87.7%
🥉 inclusionai/ling-2.6-1t — 82.6%

How does your model stack up?
https://t.co/DbnFJFig9c

[PinchBench @pinchbench]

🦞 New PinchBench result!

inception/mercury-2 scored 23.93% (35.18/147)
⏱️ 4341s

Provider: Inception
Gateway: OpenRouter

Result → https://t.co/axnWfXq9fq
See the full leaderboard → https://t.co/DbnFJFig9c

[PinchBench @pinchbench]

🦞 New PinchBench result!

inception/mercury-2 scored 25.85% (38.0/147)
⏱️ 3623s

Provider: Inception
Gateway: OpenRouter

Result → https://t.co/egcPpW2PQg
See the full leaderboard → https://t.co/DbnFJFig9c

[PinchBench @pinchbench]

🦞 New PinchBench result!

inception/mercury-2 scored 25.25% (37.1/147)
⏱️ 3001s

Provider: Inception
Gateway: OpenRouter

Result → https://t.co/qMBup8HyMi
See the full leaderboard → https://t.co/DbnFJFig9c

[PinchBench @pinchbench]

🦞 New PinchBench result!

inception/mercury-2 scored 25.78% (37.9/147)
⏱️ 2946s

Provider: Inception
Gateway: OpenRouter

Result → https://t.co/JukghFo1OJ
See the full leaderboard → https://t.co/DbnFJFig9c

[PinchBench @pinchbench]

🦞 New PinchBench result!

inception/mercury-2 scored 23.9% (35.14/147)
⏱️ 2859s

Provider: Inception
Gateway: OpenRouter

Result → https://t.co/BwpEpBZSoL
See the full leaderboard → https://t.co/DbnFJFig9c

[PinchBench @pinchbench]

🦞 New PinchBench result!

google/gemini-3.5-flash scored 74.14% (108.98/147)
⏱️ 10653s

Provider: Google
Gateway: OpenRouter

Result → https://t.co/XbdyPYRsgv
See the full leaderboard → https://t.co/DbnFJFig9c

[PinchBench @pinchbench]

🦞 New PinchBench result!

google/gemini-3.5-flash scored 71.05% (104.44/147)
⏱️ 9366s

Provider: Google
Gateway: OpenRouter

Result → https://t.co/RF2WuxQ29Z
See the full leaderboard → https://t.co/DbnFJFig9c

[PinchBench @pinchbench]

🦞 New PinchBench result!

Gemini 3.5 Flash scored 74.08% (108.89/147)
⏱️ 9827s

Provider: Google
Gateway: OpenRouter

Result → https://t.co/JiH6R34BRK
See the full leaderboard → https://t.co/DbnFJFig9c

[PinchBench @pinchbench]

🦞 New PinchBench result!

Gemini 3.5 Flash scored 75.34% (110.75/147)
⏱️ 9810s

Provider: Google
Gateway: OpenRouter

Result → https://t.co/i8MyC0uwIC
See the full leaderboard → https://t.co/DbnFJFig9c

[PinchBench @pinchbench]

🦞 New PinchBench result!

Gemini 3.5 Flash scored 76.27% (112.11/147)
⏱️ 9667s

Provider: Google
Gateway: OpenRouter

Result → https://t.co/lPMDX37HUy
See the full leaderboard → https://t.co/DbnFJFig9c

[PinchBench @pinchbench]

📊 PinchBench Daily Leaderboard (Best Score)

🥇 claude-opus-4.7 — 91.6%
🥈 mimo-v2.5 — 91.4%
🥉 claude-haiku-4.5 — 90.4%

How does your model stack up?
https://t.co/DbnFJFig9c

[PinchBench @pinchbench]

✅ Merged on pinchbench/api: "fix: add missing open-weight model metadata"

https://t.co/QDQBXQs1Up

[PinchBench @pinchbench]

📊 PinchBench Daily Leaderboard (Best Score)

🥇 Claude Opus 4.7 — 91.6%
🥈 Xiaomi MiMo v2.5 — 91.4%
🥉 Claude Haiku 4.5 — 90.4%

How does your model stack up?
https://t.co/DbnFJFig9c