Skip to content
See pricing

Models

  • 6 models
  • 1M tokens of context
  • 400 free messages on a new account

Uncensored

Abliterated

GLM 5.3

Z.ai

Vulnerability discovery

Intelligence Index · on the original model32
Price per 1M tokens
Input
R$10.25
Cache
R$1.55
Output
R$20.50
  • 1M context
offmem-glm-5.3-abliterated

GLM 5.3 Flash

Z.ai

Cost per task

Intelligence Index · on the original model28
Price per 1M tokens
Input
R$1.25
Cache
R$0.40
Output
R$5.15
  • 1M context
  • Vision
offmem-glm-5.3-flash-abliterated

DeepSeek V4.1 Flash

DeepSeek

Pentest: 1st of 17 at MindFort

Intelligence Index · on the original model33
Price per 1M tokens
Input
R$1.45
Cache
R$0.35
Output
R$5.15
  • 1M context
  • Vision
offmem-deepseek-v4.1-flash-abliterated

Kimi K3.1 Pro

Moonshot AI

Coming soon

Highest overall index

Intelligence Index · on the original model46
Price per 1M tokens
Input
Coming soon
Cache
Coming soon
Output
Coming soon
  • 1M context
  • Vision
offmem-kimi-k3.1-pro-abliterated

Originals

Maker weights
Original models: price per million tokens
ModelStrengthIndexInput per 1M tokensCache per 1M tokensOutput per 1M tokens
Kimi K3.1 Prooffmem-kimi-k3.1-proHighest overall index46R$2.60R$1.55R$25.65
MiniMax M3.1 Flashoffmem-minimax-m3.1-flashAgents and code29R$0.55R$0.05R$2.60
DeepSeek V4.1 Flash Fastoffmem-deepseek-v4.1-flash-fastLowest output price33R$0.55R$0.05R$1.55

Benchmarks

Intelligence Index

Average of 10 evaluations

  1. Kimi K3.1 Pro46
  2. DeepSeek V4.1 Flash33
  3. GLM 5.332
  4. MiniMax M3.1 Flash29
  5. GLM 5.3 Flash28

Artificial Analysis, independent

CyberGym

Finding real vulnerabilities

  1. DeepSeek V4.1 Flash88.1
  2. GLM 5.384.5
  3. MiniMax M3.1 Flash82.2
  4. GLM 5.3 Flash—
  5. Kimi K3.1 Pro—

Model maker

Terminal-Bench 4.0

Terminal tasks

  1. Kimi K3.1 Pro65.7
  2. GLM 5.341.9
  3. MiniMax M3.1 Flash38.1
  4. DeepSeek V4.1 Flash31.2
  5. GLM 5.3 Flash—

Model maker; GLM 5.3 by Artificial Analysis

HLE with tools

Expert-level questions

  1. Kimi K3.1 Pro70
  2. DeepSeek V4.1 Flash63.9
  3. GLM 5.362.5
  4. GLM 5.3 Flash—
  5. MiniMax M3.1 Flash—

Model maker

Measured on the original model. The uncensored version uses the same weights, without the refusal layer, and has no published measurement.

  1. Artificial Analysis, model leaderboard
  2. GLM-5.3, Z.ai
  3. DeepSeek-V4.1-Flash
  4. MiniMax-M3.1-Flash
  5. Kimi-K3.1-Pro, Moonshot AI
  6. MindFort, live pentest with 17 models (Sep 22, 2026)

Retrieved on November 7, 2026.

400 free messages
on a new account.

R$1.00 in credits. About 400 short messages on GLM 5.3 Flash.

Create account See plans