A record, not a leaderboard

Editorial standards

Abliterated API Observatory makes comparisons from published list prices and official documentation. Every source and snapshot date is shown with the relevant claim. Worked figures include token volumes, unit rates and arithmetic so readers can recalculate them. Capabilities that are not documented or tested remain marked as unverified.

A record, not a leaderboard

The observatory separates a catalog listing, a documented capability and a measured behavior. A listing confirms what the provider advertises at the snapshot date; it does not independently prove that an inference route is healthy. Model records include published prices and dated provider performance aggregates. Different workloads and sample sizes prevent treating these aggregates as a controlled ranking.

Prices and observed speed

Prices were checked on 11 November 2026. All token prices are in USD per million tokens. Published offers can change; check the live catalog before purchasing.

Performance figures come from the provider’s public summary for 10 Nov 2026, 00:00–23:35 UTC. The endpoint uses the current partial hour plus the preceding 23 hourly buckets. The public response does not disclose sample counts, prompt lengths or concurrency. These are provider-reported observations, not an independent benchmark, a guaranteed throughput or a same-workload comparison with competing providers.

For context, the same snapshot reports Kimi 3.1P at 406.86 token/s with 92.63% request success, DeepSeek V4.1 Flash at 151.69 token/s with 99.7% success, GLM 5.3 at 134.84 token/s with 99.29% success, and GLM 5.3 Flash at 45.84 token/s with 98.39% success. Request latency is not time to first token. We make no service-level promise from these observations.

Retention and resale scope

UncensuredAI states that it does not retain API prompts, responses or chat content on its servers. Account data is separate; connected apps have their own storage settings. Read the retention scope.

Zero retention is the provider's own policy statement, not an independent audit of the complete provider chain. Reselling is supported by the public partner documentation. Standard keys use public pricing; dedicated rates, master keys and specific availability commitments require an agreement.

Corrections and updates

These pages are dated publications. We do not automatically refresh their numbers in the background. Send a correction with the URL, the disputed claim and an authoritative source to editor@bestabliteratedapi.com. We can then verify and update the affected record.