Text model · #5 on the leaderboard
Inkling 2 Mini Uncensored
inkling-2-mini-uncensored · built on Inkling 2 Mini · 256K context
What we measured
OpenUncensored Score
75.7
#5 on the leaderboard
Intelligence
62
Base model GPQA Diamond 90.5%, HLE 33.9%
Agentic
78
Original model's Terminal-Bench 2.1 78.3%. Our tool tasks: 40% completed, 60% ran out of turns, 0% looped, 0% crashed
Non-refusal
92%
of 36 edgy requests answered. Still refuses 0% of hard-line ones.
How each number is measured is on the methodology page.
Providers and prices
| Provider | Status | Input /M | Output /M | Cached input /M | Output tok/s | First token |
|---|---|---|---|---|---|---|
| OpenUncensored GPU | online | $0.15 | $0.60 | not published | 132 | 0.6 s |
Requests go to the best provider first and fall back to the next if one fails. Live speed on real traffic: 137ms · 136 tok/s.
- You pay the price of the provider that answered, with no markup. The response says which one (
ou.providerand theX-OU-Providerheader). - Cached input is what a provider charges for prompt tokens it has seen recently. Coding agents resend the whole conversation every step, so most of their tokens are cached and this column decides the bill, more than the input price.
- By default we pick the cheapest provider among those producing at least 100 tokens per second, then fall back to the rest. Ask for the cheapest provider (
cheapest) or the fastest (fastest) instead, per request (how) or for your whole key on the credits page. How it works. - Speed columns are live averages over each provider's recent requests and change through the day. A provider without recent requests shows its median over the past week.
Call it
curl https://api.openuncensored.com/v1/chat/completions \
-H "Authorization: Bearer $OU_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "inkling-2-mini-uncensored", "messages": [{"role": "user", "content": "Hello"}]}'