Abliterated LLM models Prices, limits, and ids
Abliterated open-weight models, served without added content filters. Prices are per 1M tokens.
| Model | Modalities | Context | Input $/M | Cached $/M | Output $/M | Max output |
|---|---|---|---|---|---|---|
| Abliterated modelsRefusals removed from the weights. | ||||||
Kimi K3.1 ProBest at cyberNewkimi-k3.1-pro-uncensored | 1M | 131K | ||||
DeepSeek V4.1 Flash 1017deepseek-v4.1-flash-1017-uncensored | 1M | 131K | ||||
DeepSeek V4.1 Flash 1017 Fastdeepseek-v4.1-flash-1017-fast-uncensored | 1M | $0.24 | $0.05 | $0.90 | 131K | |
GLM-5.3glm-5.3-uncensored | 1M | $2.00 | $0.30 | $4.00 | 131K | |
GLM-5.3 Flashglm-5.3-flash-uncensored | 1M | $0.24 | $0.06 | $1.00 | 131K | |
Hy4hy4-uncensored | 1M | $1.60 | $0.30 | $3.20 | 66K | |
| Standard modelsThe original models, for a fraction of the price. | ||||||
Kimi K3.1 Prokimi-k3.1-pro | 1M | $0.30 | 131K | |||
DeepSeek V4.1 Flash 1017deepseek-v4.1-flash-1017 | 1M | 262K | ||||
Which model should I use?
| The job | Use | Why |
|---|---|---|
| Long agent runs | DeepSeek V4.1 Flash 1017 | The lowest bill of our models at the agent mix |
| Security research and red teaming | Kimi K3.1 Pro | The best model for this work |
| Video and screen recordings | Kimi K3.1 Pro | Our only model that takes video |
| Screenshots, charts and images | Hy4 | Takes images at a lower bill than Kimi K3.1 Pro |
| Hard reasoning for less | GLM-5.3 at max effort | A lower bill than Kimi K3.1 Pro |
| Batch jobs and background work | GLM-5.3 Flash | The lowest input price of our models |
| The hardest workloads | Kimi K3.1 Pro | The state-of-the-art open model |
What 1M agent tokens cost
Each square is $0.01. We assume 97% cached input, 2% uncached input, and 1% output.
- DeepSeek V4.1 Flash 1017$0.05
- DeepSeek V4.1 Flash 1017 Fast$0.06
- GLM-5.3 Flash$0.07
- Hy4$0.35
- GLM-5.3$0.37
- Kimi K3.1 Pro$0.49
More detail in our guides to coding agents, security research and abliterated AI.