- GPT-5.6 covers the price band with three tiers: Sol ($5/$30), Terra ($2.5/$15) and Luna ($1/$6);
- Anthropic reaches up-market with Fable 5 / Mythos 5 at $10/$50, while Opus 4.8 ($5/$25) matches Sol head-on.
- Traffic: ChatGPT's 5.38B monthly visits still run 5.7× Claude's 947M — but Claude is +850% YoY vs ChatGPT's −2%. Both cooled MoM in June; whether the new models re-ignite growth is July's biggest question.
- Claude's official extension leads at 10M installs; Claude's −33.8% MoM app downloads is the pullback after the growth spike.
1. GPT-5.6: one number, three names — opposite Fable 5 & Mythos 5, two naming philosophies
The biggest change isn't benchmarks — it's the naming system: the number (5.6) marks the generation, while Sol / Terra / Luna are capability tiers that can advance independently — Sol targets the highest reasoning ceiling and long-running agentic work, Terra is the balanced default, Luna the lightweight low-cost tier. Anthropic's release a month earlier sliced things differently: Claude Fable 5 is the public version with additional safety classifiers (triggering in under 5% of sessions on average, with Opus 4.8 answering as fallback); Claude Mythos 5 shares the same base model without that layer, available only to approved organizations via Project Glasswing.
Reading the two columns against each other reveals three layers: a head-on mid-flagship clash — Sol and Opus 4.8 both charge $5 input, with Opus 4.8 $5 cheaper on output; only Anthropic holds the top — Fable 5 / Mythos 5 at $10/$50 costs double Sol, and OpenAI fields nothing at the $10 tier this generation; a knife-fight at entry level — Luna ($1/$6) prices right against Haiku 4.5 ($1/$5), while Terra ($2.50/$15) undercuts Sonnet 5's standard price ($3/$15). OpenAI is carpeting the price band; Anthropic is planting a flag at the premium end — the pricing table is each company's strategy map.
2. Four surfaces: the real cards on the table
Model launches are narrative; traffic is fact. Here is each company's latest monthly data (June 2026) across web, app, browser extension, and model portfolio.
OpenAIvs
AnthropicOpenAI's base remains massive: #1 on both web and app charts, with chatgpt.com alone drawing 5.38B monthly visits — 5.7× Claude. But the trend matters: June web -3.3% MoM, app downloads -7.6%, and the 14-month curve has turned negative YoY (-2.0%). That's not collapse — it's the normal oscillation of a high plateau. The real issue is that the incremental battle (the share Gemini and Claude captured over the past year) keeps widening. GPT-5.6's timing reads less as a flex and more as a defense of the base.
For Claude, this was the month the growth king shifted gears. The good: claude.ai passed Canva to reach #3 on the global web chart (behind only ChatGPT and Gemini), with +850% YoY still the steepest curve among top products; the official browser extension crossed 10M installs, +25% MoM — ahead of every third party and 2.5× OpenAI's official extension (4M). The bad: four straight months of steep climb paused in June (953M → 947M, -0.6%), and app downloads fell -33.8% MoM. Note that downloads measure new acquisition, not active usage — a pullback after explosive growth is textbook — but the size of the drop says May's download spike (the attention around Fable 5's launch window) did not carry into June.
3. The other headline that day: a usage-limit reset
On July 10 — the day GPT-5.6 reached full availability — Anthropic announced via official channels that it had reset the 5-hour and weekly usage limits for all users across Pro / Max / Team / Enterprise. Compute is now the binding constraint for both companies.