I dont understand why there is not a gpt-6 terra?

This is all just running in circles. The models are not obviously better. The pricing fluctuates or offset by some other less-obvious metrics (availability/speed/tokens per task/dumbing down). Everyone reports different outcomes in their usage because it's all so context and user dependent. Sometimes models do some things better but become so annoying and obtuse in their other doings that it's just not worth it (like Opus with the insane code comments and Astra with its over-the-top, everything-is-a-sales-pitch style). It feels like the AI gods just turn the knobs on things like compute to get the results they want to align with the IPO to make headlines.

Anthropic it is game over when they passed OpenAI at the beginning of the year.

Now it is revenge time and OpenAI kind of is trolling Anthropic by simply going into a price war with impressive performance.

OpenAI is doing a decent job this year after they recovered. Anthropic needs to offer more payment options and be clear about token usage. The warnings I got when switching to Fable 5.1 felt like a thread. I bet more and more on OpenAI since I don’t feel robbed by them.

To me it's super interesting that OpenAI released GPT-6 Sol and Luna & Anthropic released Opus 5.5 within hours of each other.

These models are significantly cutting down the token costs by almost 40-50% as compared to their predecessors. This is exactly what people need - cutting edge intelligence at half the cost.

Wow. Luna 6 is an insane value bargain at this point. If Anthropic doesn't finally come out with their own small low cost model instead of still having haiku 4.5 they'll be history soon.

I'm trying to test them but the Visual Studio Codex Extension on WSL2 is not helping.

It seems there's a bug, shipped together with the flag that enables the new models, that doesn't allow Codex to run properly in the WSL2 sandbox.

Interesting that factual error rates have not improved a lot since the last year. Models are getting smarter but not more trustworthy. At this point I would love to have a model that's maybe not as smart but has lower factual inaccuracies and works harder. I.e don't try to find shortcuts as much and does more rigorous self checks.

Live notification in Codex:

> GPT-5.6-Sol is retiring. This conversation will automatically switch to GPT-6-Sol

I don't recall OAI retiring a model so early lol. Similar arch?

It looks like the optimal pattern is to have Astra as the orchestrator and Sol as the implementer. Same as with Fable and Opus.

The hard part for me is choosing the model and effort, that's why I always resort to Astra xhigh, but then it ends up consuming tokens so fast.

How do you decide what to pick? I mean, I do Platform work on a large monorepo with many different interconnected services, and so I always want the implementation to be "correct".