The Most Crowded Flagship Race Yet
The past six weeks delivered more frontier-model launches than the entire previous quarter. Anthropic introduced Claude Fable 5, a new tier above Opus. OpenAI answered with GPT-5.5. Google shipped Gemini 3.5 Flash at I/O 2026, and xAI lobbed a price grenade with Grok 4.3 at $1.25/$2.50 per million tokens. If you locked in your model choice in March, your cost-performance math is already out of date.
June 2026 Flagship Pricing at a Glance
| Model | Input $/1M | Output $/1M | Context |
|---|---|---|---|
| Claude Fable 5 | $10.00 | $50.00 | 1M |
| GPT-5.5 | $5.00 | $30.00 | 400K |
| Claude Opus 4.8 | $5.00 | $25.00 | 1M |
| Gemini 3.1 Pro | $2.00 | $12.00 | 1M |
| Gemini 3.5 Flash | $1.50 | $9.00 | 1M |
| DeepSeek V4 Pro | $1.74 | $3.48 | 1M |
| Grok 4.3 | $1.25 | $2.50 | 1M |
Two things stand out. First, the flagship price spread is now 20x on output tokens — $2.50 for Grok 4.3 versus $50 for Fable 5. Second, 1M-token context windows are now table stakes: every flagship except GPT-5.5 offers one at standard pricing.
Who Should Pay for the Premium Tier?
Claude Fable 5 doubles Opus pricing, and Anthropic is betting that long-horizon agentic work justifies it: fewer retries, fewer human corrections, and better end-to-end task completion can easily outweigh a 2x per-token premium. For chat and summarization workloads, though, Opus 4.8 at $5/$25 — or even Sonnet 4.6 at $3/$15 — remains the rational pick.
GPT-5.5's headline number is less interesting than its 90% cached-input discount ($0.50 per million cached tokens). Agent loops that re-read large system prompts every turn will see effective input costs far below the sticker price.
The Budget Story: Grok and DeepSeek
Grok 4.3 is the most aggressive flagship pricing we have ever tracked — roughly a tenth of typical flagship output costs with a 1M context window. DeepSeek V4 continues the lab's cache-first strategy: cache hits drop input pricing by up to 50x, making V4 Flash ($0.14/$0.28) extraordinary value for high-volume pipelines.
What to Do This Month
Re-run your workload numbers. Our cost calculator and agent cost calculator are updated with all June 2026 prices, and the comparison tool now includes Fable 5, GPT-5.5, Gemini 3.5 Flash, Grok 4.3, and DeepSeek V4. A model decision made 90 days ago could be costing you 2-5x more than necessary today.