Model updates: July 2026
Three big releases in three weeks — and the cheap tiers, not the flagships, are the real story.
Updated 31 Jul 2026 fast-moving — check a current source
Model names, prices and dates verified against provider announcements in late July 2026. All of it moves — check the source before relying on a figure.
July brought a major release from each of the three big providers inside three weeks. That pace is itself the news: releases have shifted from occasional landmark launches to steady increments on cost, speed and efficiency.
What actually shipped
- OpenAI’s GPT-5.6 family reached general availability on 9 July, in three tiers — Luna, Terra and Sol, cheapest to most capable. All three share a one-million-token context window and a February 2026 knowledge cutoff.
- Anthropic’s Claude Opus 5 landed on 24 July, aimed at long-running agent work, following Claude Sonnet 5 on 30 June.
- Google’s Gemini 3.6 Flash and 3.5 Flash-Lite went generally available on 21 July, both pitched on token efficiency and cost rather than raw capability. Gemini 3.5 Pro was not part of the release — Google has said it is still testing with partners.
The pattern worth noticing
The headline models are not where the movement is. Every provider now sells a cheap, fast tier — Luna, Flash-Lite, Haiku — and those tiers are getting good enough for a large share of everyday work. OpenAI’s published API prices span a fivefold range across the GPT-5.6 family, and Google positioned 3.6 Flash explicitly on using fewer tokens than the model it replaces.
The second pattern: a million-token context window is becoming an unremarkable baseline rather than a premium feature. “Paste in the whole document” is now a reasonable default across all three providers — though longer context still costs more and reads less carefully than a focused prompt.
A benchmark score tells you which model won a test. It does not tell you which model is right for your work. If a cheaper tier does your task acceptably, the capability gap at the top of the range is not yours to pay for.
What this means in practice
For most people, nothing urgent. The assistant you already use got better in July. The case for switching is weak unless you are hitting a specific limit — cost at volume, context length, or a capability gap you can actually name.
If you build on the APIs, the cheap tiers are worth re-testing. Work that was uneconomic on last year’s pricing may not be now. For how the assistants compare on real tasks, see the writing and coding comparisons.
Sources
Everything above was checked against these on 31 Jul 2026. Providers change things without notice — if a detail matters to a decision, follow the link.