AI

Recursive Flint - the "Qwen3.8 35B" we've all been waiting for?

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 22/10/2026, 09:47:26 UTC.

Briefings

Recursive announced a new coding agent model named Recursive Flint, featuring 40B parameters with 2.5B active, achieving top scores on TerminalBench 2.1 and SWE-bench Pro v2. The model utilizes 3:1 SWA:full attention and supports 262k context windows. While it aims to surpass Qwen3.6-35B-A3B, initial reports suggest Ornith-1.5 remains the standard in this class, though early implementation glitches may affect performance.

Key points

• Recursive Flint is a coding agent model with 40B total parameters and 2.5B active parameters.
• The model achieved top scores on TerminalBench 2.1 and SWE-bench Pro v2, potentially surpassing Qwen3.6-35B-A3B.

Context

Recursive's blog post and WIRED Magazine coverage highlight the model's performance in coding tasks. The release aims to improve coding capabilities for GPU-poor environments, though initial reports suggest Ornith-1.5 remains the standard in this class of models.

Limitations

Initial reports suggest Ornith-1.5 remains the standard in this class of models. Early implementation glitches may affect performance, and benchmarks to come in the next few days will determine the final results.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 22/10/2026, 09:48:28 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeie2ev24nkg3bc5bdi6k6m5sfjmyogt6ioprho43gvzemcaowh62ma