Coinbase fine-tunes Qwen3.5-9B model for fraud prevention, beating frontier models
Coinbase said it fine-tuned the open-weight Qwen3.5-9B model on proprietary transaction data for fraud detection, and the model outperformed Claude Opus 4.5 in its tests. The experiment also included a three-part onramp setup, where Coinbase said the model produced better fraud scores and lower serving latency, and one report said the largest gain came in spotting the monetary value of fraudulent transactions.
What this rests on
What it means
Each sentence links inside to the piece that supports it.
Users and Coinbase customers see a fraud-detection setup that is more specific than a general-purpose model, with Coinbase saying the tuned model handled the task better in its tests. The broader AI and crypto sector also gets a live example of a smaller, specialized model being used for a narrow financial job rather than a general frontier system.
Market reaction
How the story was built
Every piece in order, with the time its news medium declares — if it is not believable we use ours and say so — and what it added to what was already known at that point. Arriving first is not the same as adding something.
Crypto Briefing
scoop
AMBCrypto
echo
Where this sits in the Bitcoin series
| When | What happened | Figure |
|---|---|---|
| 1.0 years ago | BTC a 1.152e+05 $ high | today it sits -29.2 % from there |
| 4 months ago | BTC a 5.857e+04 $ low | today it sits +39.3 % from there |
From what we have stored, starting 1.0 years ago.