# GPT-5.4 ships, Karpathy's autoresearch, and Qwen3.5 goes small

> Week of 2 Mar 2026: OpenAI releases GPT-5.4, Google adds Flash-Lite, LTX-2.3 opens up, Qwen3.5 small models arrive, and Karpathy open-sources autoresearch.

Published 13 Oct 2026. Coverage: 2 Mar 2026 to 8 Mar 2026. By Ali Asif.

## Items

1. **GPT-5.4** (OpenAI, Frontier): OpenAI's new flagship, folding its coding strengths into the main model with stronger computer-use and tool-handling. [Primary source](https://developers.openai.com/api/docs/models/gpt-5.4)
2. **Gemini 3.1 Flash-Lite** (Google DeepMind, Frontier): Google's cheapest Gemini 3 tier, built for high-volume work where cost per call is the constraint. [Primary source](https://deepmind.google/models/model-cards/gemini-3-1-flash-lite/)
3. **Qwen3.5 small series (0.8B to 9B)** (Alibaba Qwen, Open weights): Open small models, from 0.8B to 9B, meant for phones, laptops and edge devices. [Primary source](https://huggingface.co/Qwen/Qwen3.5-9B)
4. **LTX-2.3** (Lightricks, Image, video, world): An open video-and-audio model with a rebuilt VAE for sharper detail and native vertical video. [Primary source](https://huggingface.co/Lightricks/LTX-2.3)
5. **autoresearch** (Andrej Karpathy, Agents and coding): A small repo where an AI agent runs short training experiments on one GPU, keeps what improves, and discards the rest. [Primary source](https://github.com/karpathy/autoresearch)

OpenAI had the big headline, but the repo that stayed in my head came from a single person.

## GPT-5.4

OpenAI released GPT-5.4 on 5 March. The shape of the release matters: the coding strengths of the Codex line are folded into the main model, along with better computer use and tool handling. The separate-models era is easing off, at least at the top end.

For anyone building agents, that simplifies life. One model that writes code, drives a browser and calls tools means fewer hand-offs between systems, and fewer places for a task to fall over. I'll run it against the same internal tasks I used on Opus 4.6 and keep the results in a table, because my memory of "which felt better" is not evidence.

## Cheap and small

Google released Gemini 3.1 Flash-Lite, the lowest-cost Gemini 3 tier. Alibaba filled out the Qwen3.5 family with small open models from 0.8B to 9B parameters. These two releases serve the same need from opposite sides: lots of cheap calls, either through an API or on your own device.

## Video

Lightricks released LTX-2.3, an update to its open video-and-audio model. It rebuilds the VAE for finer detail, retrains the audio path, and supports true portrait video, which matters for anyone producing vertical content.

## Karpathy's autoresearch

On 7 March Andrej Karpathy put autoresearch on GitHub. An agent reads instructions, edits the training code for a small model, runs a fixed five-minute experiment, checks the result, and keeps or throws away the change. Then it repeats. It runs on a single GPU.

What I like is the discipline. One file the agent may edit, one metric, a fixed time budget. Most agent failures I have seen come from vague goals and unlimited scope. This is the opposite.

One detail on the GPT-5.4 launch that deserves a second look is how many vendors now talk about computer use as a core skill instead of a demo. If a model can reliably operate a browser and a spreadsheet, a lot of the glue code I write to connect systems becomes unnecessary. I treat that as a promising direction, and I also treat it as something to test on a dummy account first, since a model clicking in the wrong place has real consequences.

**What I'd watch next:** whether people apply the autoresearch loop to ordinary business metrics, not only model training.

## My take

Autoresearch is a tiny repo with a big idea, and it is how I think about running agents on my own work: a fixed loop, a clear score, and the agent keeps only changes that improve it. The same pattern fits supplier-quote checking or BOQ validation. Pick one metric, let the agent iterate overnight, and review the winners in the morning.

## Sources

- https://developers.openai.com/api/docs/models/gpt-5.4
- https://deepmind.google/models/model-cards/gemini-3-1-flash-lite/
- https://huggingface.co/Qwen/Qwen3.5-9B
- https://huggingface.co/Lightricks/LTX-2.3
- https://github.com/karpathy/autoresearch

Canonical: https://heyaliasif.com/ai/2026-03-02-week-in-ai