On November 12 at Baidu World 2026, Robin Li launched Token Exchange, a model marketplace he compared to a stock exchange for AI models. It brings many model service providers together so developers integrate one API, compare models and providers on quality, latency and price, and switch between them. To counter resellers that swap models or inflate specs, Baidu vets provider credentials and runs model fingerprint comparisons plus capability, performance and availability tests before listing, then spot-checks providers afterward and keeps a public credit rating and record for each. Baidu says that one month in, the platform lists over 100 model resources and handles more than 50M calls and 700B tokens a day.
Key Takeaways
- ✓Positioning: a neutral model brokerage platform that Robin Li calls an exchange for AI models (Baidu post)
- ✓Scale (per Baidu): one month in, 100+ model resources, 50M+ calls and 700B+ tokens a day
- ✓Listing checks: provider vetting, model fingerprint comparison and capability, performance and availability tests, aimed at model swapping and inflated specs
- ✓Ongoing oversight: regular spot checks, continuous quality monitoring and provider credit ratings covering service quality, delivery, compliance and contract performance, with alerts to users when a rating drops
- ✓Integration: one API to compare quality, latency and cost across providers and switch freely; the product page still shows older figures (70+ resources, 200B tokens a day)
Key Decision Metrics at a Glance
Turn your technical choice into a development budget
Compare 40 dev plans & simulate token costs vs $20/mo subscriptions
Project Links & Resources
Direct AccessIn-Depth Technical Analysis
Background
At Baidu World 2026, Robin Li said that in the first nine months of the year more than one new model shipped almost every day, demand for model calls exploded, and a wave of token resellers appeared. Those resellers take no responsibility for quality, stability or latency, and often cannot prove they serve the model they advertise. Prices for the same model vary widely and sources are murky, which makes enterprise model selection risky.
Architecture highlights
Token Exchange is a brokerage platform under Baidu AI Cloud, not just a channel for Baidu's own models. Many providers list model resources; users integrate one API to compare quality, latency and cost and switch at any time. Trust works in three layers: before listing, provider vetting, model fingerprint comparison and capability, performance and availability tests; after listing, regular spot checks and continuous quality monitoring; and provider credit ratings on service quality, delivery, compliance and contract performance, kept as a public record, with alerts to users when a rating drops.
Benchmarks
As a platform, it has no model benchmark. Baidu's figures: after one month, more than 100 model resources, over 50M calls and over 700B tokens a day. Note that the Baidu AI Cloud product page still shows 70+ resources, 200B+ tokens and 60M+ calls a day, which does not match the launch figures, so treat the numbers as Baidu's and wait for an official update. Baidu has not published how accurate its fingerprint checks are.
Getting started
If you route calls to several models through third-party resellers, Token Exchange is a verified alternative: enable it on Baidu AI Cloud, run the same task set across providers through the unified API, compare latency and success rates on the live rankings before choosing, and watch provider credit labels. For critical workloads, keep a direct official API as a fallback while you see whether the spot-check and rating system proves transparent.
Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.