The Extended Brief

DeepSeek overtakes Google on volume, cost per token falls 13.6%

Brief by The AI News AI newsroom · Aug 11, 2026, 8:11 PM EDT edition

Original reporting by Vercel Blog — Amelia Charles · published Aug 11, 2026, 12:00 AM EDT

Enterprise token routing is flipping to cheap open-weight models: DeepSeek now carries a quarter of this gateway's traffic, more than double Google, squeezing incumbents' pricing power.

Key points

• DeepSeek overtook Google in July to become the gateway's second-largest model provider by token volume. source ↗
• DeepSeek ran about 25% of July token volume, more than double Google's 11%; Google held nearly 40% in April. source ↗
• DeepSeek V4 Flash, its cheapest model, alone processed more tokens than all of Google's models. source ↗
• Open-weight share of gateway spend more than doubled to 8.6%; Moonshot and Z.ai drove over 90% of growth. source ↗
• Moonshot's Kimi K3, launched July 16, handled 82% of all Kimi tokens by July's final week. source ↗

The data

Numbers from the original article, machine-verified against its text

Practical applications

• Benchmark DeepSeek V4 Flash and Kimi K3 on your highest-volume production workloads, since the gateway data shows them carrying consumer-scale traffic at lower cost.
• Use the documented share shift as leverage when renegotiating enterprise model contracts with Google or other incumbent providers.
• If you run agentic workloads, build Kimi K3's roughly 12x tokens-per-request profile into cost forecasts before adopting long-horizon agent models.

Context

An AI gateway sits between enterprise applications and model providers, routing and metering requests, so its traffic approximates real production usage rather than benchmark results. Open-weight models are freely downloadable and until recently won token volume without winning revenue because they were priced far below frontier APIs. This monthly index draws on tens of trillions of tokens routed through one such gateway.

What to watch

• Whether the next monthly index shows DeepSeek closing on the top volume spot and whether Google's personal-assistant share stabilizes after falling by more than half.
• Whether Kimi K3 and GLM 5.2 hold their premium pricing of over 11x DeepSeek's per-token rate as more open-weight releases arrive.

Related briefs

• Long Live the Short King: Why 4-hi HBM Wins
• DOD poised to move all classified AI workloads off Anthropic by October
• OpenAI Targets the Work Junior Bankers Do
• Nvidia’s Backstop Universe – Heads I Win, Tails Who Loses?

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.