Helix AI on Tuesday released Reasoner 2, its latest open-weight reasoning model, positioning it as a lower-cost alternative for math, coding, and multi-step analysis workloads.
The 70-billion-parameter model ships with a 128,000-token context window and was trained with a reinforcement-learning pipeline focused on verifiable reasoning tasks. In evaluations shared with the release, Reasoner 2 solved 78 percent of a held-out set of competition math problems and passed 64 percent of a SWE-style coding benchmark, figures that place it roughly in line with larger proprietary systems on structured tasks.
Weights are available under an Apache 2.0 license, and the company is offering API access at $1.20 per million input tokens and $4.80 per million output tokens. A distilled 8-billion-parameter variant aimed at on-device and edge deployment is expected next quarter.
Independent researchers cautioned that vendor-reported benchmarks should be treated as directional until third-party evaluations land, but noted the release continues a clear trend: open models are closing the gap on closed systems for reasoning-heavy work while undercutting them sharply on price.
Reasoner 2 is available now through the company’s developer platform and major model hubs.
