Ornith releases open-weight Ornith-1.5-122B under MIT license, runs on a single DGX

The Ornith team released the open-source model Ornith-1.5-122B, opening its 122B-parameter weights under the MIT license, pitching "no trade-off between strong and fast": it claims SOTA among open models of the same scale while remaining efficient enough to run on a single NVIDIA DGX machine, with no large inference cluster required.

Confirmed

• Ornith-1.5-122B was released with weights open-sourced under the MIT license, 122B parameters
• Official benchmark scores: 84.5% (0.845) on coding benchmark SWE-bench Verified, 81.3% on Terminal-Bench 2.1, and more
• The team claims it runs on a single NVIDIA DGX machine while delivering frontier-level capability

Not Yet Confirmed

• All benchmark numbers and the "SOTA at this scale" claim come from official promotion (original poster @ornith); most posts in this batch are reposts, with no independent reproduction or third-party evaluation yet

Why It Matters

If the official results hold up under independent verification, the model's fully open MIT license and single-machine deployability would significantly lower the barrier to frontier coding/agent capabilities, putting direct competitive pressure on both closed-source APIs and open-source setups requiring inference clusters. A wave of reposts appeared October 20–21, indicating strong community interest.

2026-10-20 ~ 2026-10-21 · 6 related posts

• Open-weights Ornith-1.5-122B launches with 84.5% SWE-bench Verified, runs on a single DGX card — airesearch12 · 2026-10-20
• Open-weights Ornith-1.5-122B debuts with 84.5% SWE-bench Verified, runs on a single DGX card — _akhaliq · 2026-10-21

4 near-duplicate retellings: hey_abusiddik · alifcoder · iamfakhrealam · aftahi_ai