The Extended Brief
Exclusive: Paying for frontier AI models buys 4-month head start at 5x the cost
Brief by The AI News AI newsroom · Sep 15, 2026, 9:21 AM EDT edition
Original reporting by Ars Technica AI — Jeremy Hsu · published Sep 15, 2026, 8:00 AM EDT
Open models now trail frontier models by only 4.4 months, so teams paying frontier prices for routine work are likely overpaying.
Key points
- Mozilla's report finds open-weights models now trail closed frontier models by just 4.4 months in performance. source ↗
- Moonshot AI's Kimi K3 scores three points behind Anthropic's Fable 5 on the Artificial Analysis Intelligence Index. source ↗
- Kimi K3 costs 30 percent of what Fable 5 does, per the report. source ↗
- Mozilla CTO Raffi Krikorian says closed models earn their premium in expert professional work, high-intensity retrieval, and long context. source ↗
- The report, published September 15, recommends most organizations default to open models for the majority of their work. source ↗
The data
4.4 months
Performance lag of the best open-weights models behind closed frontier models
From Mozilla's State of Open Source AI report, published September 15.
30%
Kimi K3's cost relative to Anthropic's Fable 5
K3 scores three points behind Fable 5 on the Artificial Analysis Intelligence Index.
Numbers from the original article, machine-verified against its text
Practical applications
- Benchmark a leading open model such as Kimi K3 against your current frontier model on routine workloads before your next contract renewal.
- Confine closed-model spend to the three workloads Mozilla's CTO says justify it: expert professional work, high-intensity retrieval, and long context.
- Pull Mozilla's State of Open Source AI report for its workload breakdown before setting next year's model budget.
Context
Open-weights models can be downloaded and run by anyone, while closed frontier models are reachable only through paid vendor services. The report identifies Chinese companies, including Moonshot AI, as the source of the best open-weights models. The Artificial Analysis Intelligence Index is a composite benchmark score used to compare capability across both camps.
What to watch
- Whether the 4.4-month gap keeps shrinking in future Artificial Analysis index updates will show if the closed premium erodes further.
- Watch for frontier vendors cutting prices or sharpening differentiation on long-context and retrieval workloads in response.
Related briefs
- Anthropic Makes $13.7 Compute Deal With Trump-Linked Rum Group
- Apple's Siri AI Can Be Swapped Out for Claude, ChatGPT, Code Shows
- Inside ‘Project Lily’: The Humans Reading Your ChatGPT Chats
- Long Live the Short King: Why 4-hi HBM Wins
Editorial score 4.0 / 5 · significance 4.0 · novelty 4.0 · edge 4.0 · perspective 4.0
Desks: Business · Engineering
Topics: Pricing & economics · Open-source AI · 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.