5 releases, +0.3%: open-weight coding hit a wall

Addy Crezee 1 Jun 2026 1 min read

Man shocking a CRT monitor with defibrillator paddles in a dark room, sparks flying, illustrating stalled open-weight coding

swe-bench pro across open-weight models hasn't moved in 10 weeks

mar 27th - jun 1st. 5 major open-weight releases. net progress on swe-bench pro: +0.3%. scores swung -5% and back within that window

looks like open-weight labs are optimizing hard for agent tasks and multimodality – but not for coding anymore

Table comparing five open-weight models on SWE-bench Pro from March to June 2026, scores clustered between 55.4 and 59

Introducing MiniMax M3: The First Open-Weights Model to Combine Three Frontier Capabilities

- Coding & Agentic Frontier: 59.0% SWE-Bench Pro, 66.0% Terminal Bench 2.1, 34.8% SWE-fficiency, 28.8% KernelBench Hard, 74.2% MCP Atlas
- MiniMax Sparse Attention scales context to 1M
-… pic.twitter.com/TF891iJukF
— MiniMax (official) (@MiniMax_AI) June 1, 2026

deepmind open-sources science skills: ai agents meet 30+ databases

google is quietly making a strategic move in the war for scientific talent deepmind just open-sourced "science skills" – a toolkit that lets ai agents natively talk to 30+

4 Jun 2026 1 min read

Surgeon cracking open a chip labeled LLM Backbone with Vision and Audio modules wired in, monitor showing 12B

google launches gemma 4 12b – nearly matches the 26b model on benchmarks, sometimes beats it, at less than half the memory footprint

what changed under the hood: • vision. replaced the encoder with a lightweight embedding module (single matrix multiply + positional embedding + normalization). the llm backbone now handles visual processing directly • audio. encoder

3 Jun 2026 1 min read

Man in gray suit presents llama figurine under glass dome labeled NEW to skeptical exec in dark graffiti-tagged boardroom

meta's chief ai officer alexandr wang is closing out his first year – and it's not going well

why he's being criticised: • leadership style called chaotic and impatient • accused of hyping small, step-by-step improvements as if they were major breakthroughs • talent leaks – a former apple exec

3 Jun 2026 1 min read

Stay in the loop

Read next

deepmind open-sources science skills: ai agents meet 30+ databases

google launches gemma 4 12b – nearly matches the 26b model on benchmarks, sometimes beats it, at less than half the memory footprint

meta's chief ai officer alexandr wang is closing out his first year – and it's not going well