Ornith 1.5 open models match Claude Opus 4.8 on coding benchmarks

An open-source model family just posted numbers that rival one of the most capable closed models in the world — and it's available under an MIT license.

Ornith 1.5, released today by the Ornith AI team, is a three-model family spanning 9 billion dense parameters up to a 397 billion mixture-of-experts flagship. The big headline: the 397B MoE scores 86.1 on Terminal-Bench 2.1 (Terminus-2) and 86.0 on SWE-bench Verified, placing it on par with Claude Opus 4.8's 85.0 and 85.8 respectively. It also outperforms GLM-5.2 (753B parameters) and DeepSeek-V4-Flash-0731 (284B) across multiple coding benchmarks, despite being smaller than both. The 9B dense model is no slouch either, hitting 70.6 on SWE-bench Verified and 47.5 on SWE-bench Pro — competitive with models several times its size.

What makes Ornith 1.5 technically interesting is its training approach. Rather than relying on human-curated tasks and manually designed evaluation harnesses, the team extended what they call "self-scaffolding" into a full self-improvement loop. The system jointly optimizes three things: generating its own training tasks, constructing scaffolds to solve them, and running reinforcement learning on the resulting rollouts. In other words, the model teaches itself by creating problems and figuring out how to solve them — no human supervision needed for the task design. All three models are built on the Qwen 3.5 architecture and carry an MIT license, with weights available on HuggingFace alongside quantized GGUF variants for local deployment.

The open-model landscape keeps tightening. Just this week we covered GLM-5.3 tying Kimi K3 at the top of the Artificial Analysis Intelligence Index — now Ornith 1.5 is posting numbers that match frontier closed models on coding tasks. The gap between what you can run locally and what you have to pay an API for is shrinking fast.

What to watch: independent evaluations of the self-improvement training loop — if the approach generalizes, it could reshape how open models are built.

Is self-improvement the training paradigm that finally closes the open-closed gap, or will frontier labs pull ahead with the next generation? Tell us in the comments.

AI 101 — How do AI video generators work?

An AI video generator is a model that starts with a grid of random noise and repeatedly refines it into a moving picture that matches your prompt — the same trick image generators use, stretched across time. Nothing is filmed, and nothing is stitched together from clips the model has seen. The pixels in the finished video were invented, frame by frame, all at once. Why it matters right now Kling 4.0, Kuaishou's flagship video model, went fully live on October 15 with up to 30 seconds of vid

Houmo moves compute inside memory — and A20 enters production

Beijing chipmaker Houmo used its partner conference this week to lay out two compute-in-memory routes at once — and to say the first chip built on the newer one is already in production. Houmo's A20, its first chip built on a 3D compute-in-memory architecture, has entered production. The part supports up to 16 chips fully interconnected, claims a 50% improvement in compute utilization over the previous generation, and covers models from 3B to 30B parameters, according to co-founder and product

Kling 4.0 goes live with 30-second native video at 10-bit HDR

Kuaishou's Kling AI has spent two years becoming the video model people actually finish projects with. Its new flagship, Kling 4.0, went fully live on October 15 — and this round is less about prettier frames than about the two things that make a generated shot usable in a real edit: length and control. Kling 4.0 generates up to 30 seconds in a single native pass, outputs 10-bit HDR at 4K, 1080p and 21:9, and takes up to 15 reference assets in one task. That budget splits across ten images, fi

Brain organoids classified Braille at 83% — in a trio, not alone

Two stories today: human brain tissue on a chip decoded a tactile signal, and the data-centre company behind some of the largest AI contracts in Europe turns out to be six months behind on its flagship site. Human forebrain organoids grown on a microelectrode array classified Braille letters with 61 percent accuracy on their own and 83 percent when the responses of three organoids were combined, in a peer-reviewed study from the University of Bristol. The team took event-based tactile readings

Read more

AI 101 — How do AI video generators work?

Houmo moves compute inside memory — and A20 enters production

Kling 4.0 goes live with 30-second native video at 10-bit HDR

Brain organoids classified Braille at 83% — in a trio, not alone