The frontier moved to the mid-tier — Sonnet 5 and the fight for coding at scale

When a lab puts its mid-tier model at the frontier for coding and agents, it's telling you where the money is. The flagship proves capability; the tier below it is where the revenue runs.

Anthropic pitched Sonnet 5 — its mid-tier line — as frontier-grade on coding, agents, and professional work at scale. That is a segmentation move as much as a capability one: the high-throughput enterprise workloads should be served by a model priced to run them at volume, not by the most expensive tier.

Coding is the battleground because coding is the revenue

Coding and agentic work are the highest-retention, highest-value enterprise AI use cases of the cycle — the ones that produce artifacts teams keep using. A model tuned and priced for exactly those workloads is a revenue strategy wearing a capability announcement's clothes.

There is no single frontier anymore

The proof is in the benchmarks. OpenAI tops science reasoning while Anthropic tops real-world software engineering — the crown has split by domain. "Frontier" is fragmenting from a single flagship into a lineup where each tier and each lab claims the frontier for its price point and its specialty.

For buyers the takeaway is to stop shopping for the best model and start routing to the best model per task. The durable decision is the gateway, not the model behind it — because the model behind it changes every few weeks.

Anthropic — Newsroom → · David Veksler — Frontier AI labs list: companies, models and strategy (2026) →