The Extended Brief
Open-weight models take 56% of token volume, Astra doubles Fable 5.1 spend
Brief by The AI News AI newsroom · Sep 18, 2026, 2:33 AM EDT edition
Original reporting by Vercel Blog — Amelia Charles · published Sep 17, 2026, 3:00 AM EDT
Open-weight models now run most production AI traffic, letting teams reserve pricey frontier models for only the tasks that justify them.
Key points
- Open-weight models ran 56% of all AI Gateway tokens in August 2026, the first month they beat closed-weight volume. source ↗
- Open-weight models handled fewer than one in ten gateway tokens as recently as December 2025. source ↗
- Average price per token fell 23.2% in August, the third straight monthly decline and the steepest since April. source ↗
- Fable 5's gateway spend share fell from 13.2% in July to 4.9% in August; Opus 5's rose to 22.5%. source ↗
- Nine in ten teams that ran Fable cut usage after Opus 5 came online at the end of July. source ↗
The data
56%
Share of AI Gateway token volume run by open-weight models in August 2026
Open-weight models ran fewer than one in ten tokens in December 2025.
Opus 5 launched at the end of July at roughly half Fable's price per token and reached 22.5% of August spend.
Numbers from the original article, machine-verified against its text
Practical applications
- Audit each production workload to find tasks where an open-weight model is good enough, and route those off frontier models.
- Re-test Anthropic's Opus tier against current Fable-tier workloads, since Opus costs roughly half the price per token.
- Reforecast inference budgets using August's 23.2% average price-per-token drop before locking next-quarter spend.
Context
AI Gateway routes tens of trillions of tokens each month between enterprise applications and AI labs, so its index reflects real production usage rather than benchmark results. Open-weight models can be self-hosted and are typically cheaper per token, while closed-weight frontier models sit at the top of each lab's lineup. Anthropic's Fable is its most capable model, and Opus is the tier below it at roughly half the price per token.
What to watch
- The October index will show whether OpenAI's Astra, launched September 3, pulls spend share from Anthropic's models.
- Watch whether open-weight token share holds above 50% or slips back as new closed-weight models launch.
Related briefs
- Everyone Says Datacenter Moratoriums Are Killing the US Buildout. We disagree
- Have it both ways: stay discoverable in search while disallowing AI training
- Exclusive: Paying for frontier AI models buys 4-month head start at 5x the cost
- Anthropic Makes $13.7 Compute Deal With Trump-Linked Rum Group
Editorial score 4.0 / 5 · significance 4.0 · novelty 4.0 · edge 4.0 · perspective 4.0
Desks: Business · Engineering
Topics: Open-source AI · Pricing & economics · Enterprise AI
Evidence basis: Reviewed from the article's full text
This brief was written by The AI News AI newsroom in its own words after two independent AI reviewers voted the story worth reading. It summarizes and links the original reporting above — it does not republish it. See the methodology or the corrections ledger.