State of AI API Costs: June 2026

Updated 2026-08-24: the mid-tier table below carried Claude Sonnet 5 as $3.00/$15.00 with "$2 intro" and "$10 intro" beside it, and the section above it told teams to lock in workloads before the introductory rate expired on August 31, 2026. Anthropic cancelled that increase. $2.00/$10.00 is Claude Sonnet 5's standard price, and it is what the model billed at on June 30 as well, so the table now shows it plainly. The June events and every other June price are as published.

June 2026: two Anthropic launches, one Google cut

June was Anthropic's month. Two model launches changed prices at both ends of the spectrum.

Key events:

• -Jun 1: OpenAI cut GPT 5.5 input from $6.00 to $5.00/MTok (17% reduction)
• -Jun 9: Anthropic launched Claude Fable 5 at $10.00/$50.00 per MTok
• -Jun 10: Google cut Gemini 3.5 Flash output from $0.75 to $0.60/MTok
• -Jun 30: Anthropic launched Claude Sonnet 5 at a $3.00/$15.00 list rate, billing at $2.00/$10.00 per MTok

Claude Fable 5: a new ceiling

Fable 5 is the most expensive model in the registry at $10.00 input and $50.00 output. It also holds the highest reasoning confidence at 98/100. The 1M context window matches Gemini and exceeds Opus at 200K.

For teams that require peak reasoning quality, Fable 5 sets a new benchmark. For everyone else, it widened the gap between flagship and mid-tier pricing.

Claude Sonnet 5: the mid-tier upgrade

Claude Sonnet 5 replaces Claude Sonnet 4.6 as Anthropic's primary mid-tier model, scoring higher on every task type while billing at $2.00/$10.00 against Claude Sonnet 4.6's $3.00/$15.00 - a third less on both sides.

Anthropic announced that rate as introductory pricing through August 31, 2026, against a $3.00/$15.00 list rate of its own. It has since cancelled the increase, so $2.00/$10.00 is simply what the model costs - there is no window to switch inside.

GPT 5.5 reaches $5.00 parity

OpenAI's June 1 cut brought GPT 5.5 input from $6.00 to $5.00/MTok. That matches Claude Opus 4.8 exactly. The flagship tier now clusters at $5.00 input across providers.

Output pricing still differs: Opus at $25.00 versus GPT 5.5 at $30.00 per MTok.

The June pricing snapshot

Flagship tier (June 30):

| Model | Input/MTok | Output/MTok | |-------|-----------|-------------| | Claude Fable 5 | $10.00 | $50.00 | | GPT 5.5 | $5.00 | $30.00 | | Claude Opus 4.8 | $5.00 | $25.00 |

Mid tier (June 30):

| Model | Input/MTok | Output/MTok | |-------|-----------|-------------| | Claude Sonnet 5 | $2.00 | $10.00 | | Claude Sonnet 4.6 | $3.00 | $15.00 | | GPT 5.4 | $2.50 | $15.00 | | Gemini 3.1 Pro | $2.00 | $12.00 |

Fast tier (June 30):

| Model | Input/MTok | Output/MTok | |-------|-----------|-------------| | Gemini 3.5 Flash-Lite | $0.30 | $2.50 | | Gemini 3 Flash | $0.50 | $3.00 | | Claude Haiku 4.5 | $1.00 | $5.00 | | Gemini 3.5 Flash | $1.50 | $9.00 |

What to watch in July

OpenAI's GPT 5.6 family is expected to launch soon. Three tiers would give OpenAI a complete flagship, mid, and fast lineup to match Anthropic and Google.

The mid tier is now the most competitive segment. Five models from three providers sit between $2.00 and $3.00/MTok input.

See how June's changes affect your spend at /teardown, or start tracking with a free dashboard.