Claude & GPT models restored.

All models restored.

Giveaway over, winners have been messaged and credited 50M tokens each.

Will do another giveaway soon.

Added DeepSeek v4.1 Flash (deepseek-v4.1-flash)
$0.02/M tokens, cached tokens $0.005/M
(Price may increase in the future, use it while it's cheap)

This model scores better on some benchmarks (cybersecurity) than GLM 5.3, Kimi K3, and even GPT 5.6 Sol.

💰 Top-up bonuses are now live!

The more you add to your balance, the more bonus credit you receive:

$10 | +1% bonus → $10.10 balance
$25 | +3% bonus → $25.75 balance
$50 | +5% bonus → $52.50 balance
$100 | +8% bonus → $108 balance
$200 | +12% bonus → $224 balance
$500 | +16% bonus → $580 balance

We've also updated balances to represent USD instead of tokens to prevent confusion since we now have a range of models with different pricing.

Massively improved GLM 5.3 Uncensored's refusal rate across all tasks. I recommend using it in OpenCode (in terminal) where it's been tested, but it should work across all agents/IDEs. Report any issues to @ogsadly.

Added GPT 6 Sol (gpt-6-sol) - $0.2/M.
Added GPT 6 Luna (gpt-6-luna) - $0.05/M.
Added Grok 4.7 (grok-4.7) - $0.1/M.

Opus 5.5 will probably be added within a few days.

Added our smartest model yet, Claude Opus 5.5 (claude-opus-5-5) at $1/M, cached tokens $0.1/M.

Reduced price of Claude Opus 5.5 by 20%, it's now $0.8/M with $0.08/M cache. Give it a try if you haven't, it's one of our most stable models.

Added a fully uncensored model for those who need it, glm-5.3-flash-abliterated. It's fully uncensored at the model weight level, it should never refuse.

Because it needs to be self-hosted, the price is higher than the other uncensored models, at $0.2/M and $0.05/M cache. You still won't find it cheaper anywhere else, I will price match if you do.

I only recommend using this model if the other uncensored models don't work for you, due to the price. The other uncensored models remain usable.

Reduced glm-5.3-flash-abliterated pricing by 33% to $0.4/M and $0.1/M cache, also improved its reliability.

Added back Kimi-K3 - $0.15/M, $0.03/M cache.

Added GPT-6.1-Sol at $0.1/M and $0.02/M cache, really smart model almost identical to Astra on benchmarks.

Reduced glm-5.3-flash-abliterated pricing by another 50% to $0.2/M and $0.05/M cache, try it out if you've been refused by other uncensored models this one definitely won't refuse.

glm-5.3-flash-abliterated is back online, sorry for the downtime.

Significantly improved speed and reliability on GPT 6 Sol, GPT 6.1 Sol and Opus 5.5.

Also adding GLM 5.3 Abliterated soon, smartest uncensored model with zero refusals.

Reduced GLM 5.3 and GLM 5.3 Uncensored price by -66% from $0.3/M to $0.1/M and reduced cache price to $0.03/M.