[Generative AI] Organizing the Latest on ChatGPT, Claude, and Gemini | Recent AI News
In July 2026, ChatGPT, Claude, and Gemini all released new models within half a month. However, they haven't all arrived at the same time. I have listed the current status as of July 30th, along with the dates.
This is a memo not for memorizing names, but for deciding how you want to use them.
01 | These three moved in July
🔘 ChatGPT: GPT-5.6 (Sol / Terra / Luna) — Released to the public on July 9th
🔘 Gemini: Gemini 3.6 Flash — Officially released on July 21st
🔘 Claude: Claude Opus 5 — Released on July 24th
In the half-month from the 9th to the 24th, all three companies had releases. The weight of the content differs for each.
02 | ChatGPT: GPT-5.6 has been split into three
On July 9th, GPT-5.6 was released to the public. Instead of one single latest model, it is divided into three based on use case.
Sol: Top-tier. For complex reasoning and automating multi-step tasks
Terra: Mid-tier. For handling daily tasks normally
Luna: Lightweight. For low-cost, high-speed, high-volume processing
ChatGPT Work also launched on the same day. It incorporates Codex, which was for developers, into a desktop app, allowing you to delegate tasks like document creation and multi-file analysis step-by-step.
As for voice, a new model called GPT-Live was announced in early July. It uses a full-duplex method that allows you to talk while listening.
Research, studying, playing house with children. It has already become a normal part of our lives at home.
Because there is no lag in the response, the play doesn't get interrupted.
03 | Gemini: Only the lighter version moved
On July 21st, Gemini 3.6 Flash was officially released. 3.5 Flash-Lite was also released on the same day.
Flash is the model responsible for speed and cost-efficiency. This time, it is explained that they reduced the output tokens by about 17% by trimming the steps of thought. The output pricing has also decreased compared to the previous 3.5 Flash.
On the other hand, the higher-end Gemini 3.1 Pro remains in preview. As for Gemini 4, it was only announced that pre-training has begun, with no dates, architecture, or pricing released.
In other words, Google has not yet released its main contender. While it may look like the three companies are lined up, Gemini is in a different row.
04 | Claude: Opus 5 is half the price of Fable 5
On July 24, Anthropic released Claude Opus 5. The price became a topic of conversation before the performance.
The cost is $5 per million tokens for input and $25 for output. The top-tier Fable 5 is $10 and $50, so it is exactly half the price. The unit price remains the same as the previous generation Opus 4.8, while only the performance has changed.
It is treated as the default model in Claude Max and the top-tier model in Claude Pro. Even if you haven't reselected it yourself, it may have changed to Opus 5.
An 'effort' setting has also been added to toggle the depth of reasoning. The deeper you make it think, the more tokens it uses. There is no need to raise it for light tasks.
It has also been reported that it does not reach Mythos 5 in cybersecurity-related tasks. It is not a model that is the best at everything.
05 | What I learned by comparing them
Although the names and numbers are all over the place, the direction they are heading is similar.
One thing is that they have started to make you choose by tier. GPT-5.6 has three, Gemini has Flash, Pro, and Lite, and Claude has Sonnet, Opus, and Fable. The way of choosing where the latest equals the most expensive no longer fits.
Another thing is that price has become a axis of competition. Gemini 3.6 Flash had a price cut, GPT-5.6 strengthened its cheaper tier, and Opus 5 was placed at half the price of the top-tier model. It is leaning toward the question of how much to charge for the same performance.
And they are moving toward taking action. Features that touch files and apps to complete tasks, such as ChatGPT Work and Claude Cowork, have been released by each company.
However, they are not in lockstep. OpenAI and Anthropic are moving near the top tier, while Google has only updated the lighter side. It is still too early to tell where this gap will be bridged.
06 | Things to be careful about
✅ You will get confused if you choose by name. Numbers and codenames are mixed, making them impossible to remember
✅ The default model for paid plans may change without notice
✅ Preview versions are not suitable for production work
✅ It is not all price cuts. Sonnet 5 is scheduled to increase to $3 for input and $15 for output on September 1st
Pricing and the default model. Checking these two things is enough.
07 | Confirmation points
Check the name of the model you are currently using on the screen
See if the default model for your paid plan has changed
Try to see if you can move daily repetitive tasks to a cheaper model
Decide on just three tasks to leave to the top-tier model
Check the date of the information you are reading
✦ Summary
In July, three companies made moves in just half a month. However, they are not all in sync. Google has yet to release its main contender.
The focus of competition has shifted from intelligence to price. What you choose is not the model name, but the workspace for each task.
This field changes every two to three weeks. I will make sure to write the date on both notes and articles.
Thank you for reading. See you in the next article.
▼ Summary of site map and related links ▼
▼ Featured articles ▼
🦉 Thank you so much 🦉
#AI #AI活用 #AIで日常改善 #日常改善 #暮らしを整える #生活改善 #効率化 #仕事効率化 #業務改善 #働き方 #仕事術 #ライフハック #デジタル活用 #ツール活用 #プロンプト #プロンプト設計 #AIと暮らす #AIのある生活 #AI初心者 #AI学習 #学びの記録 #体験記 #気づき #考え方 #習慣化 #note #スキしてみて #深夜ラテ