Skip to content

4 Uncensored LLM model added, currently available for only pay-as-you-go users.See Models

DeepSeek V4 Pro (latest) logo

DeepSeek

DeepSeek V4 Pro (latest)

DeepSeek V4 Pro (latest) from DeepSeek, with a 1M-token context window. Call it as deepseek/deepseek-v4-pro through the OpenAI- or Anthropic-compatible endpoint with a Tokens key.

deepseek/deepseek-v4-pro
Input, per 1M tokens
৳168.75$1.35
Output, per 1M tokens
৳500.00$4.00
Context window
1M1,000,000 tokens
Available on
Pay as you go

Measured speed and reliability

How DeepSeek V4 Pro (latest) has performed for real requests, timed at our gateway in Dhaka.

Not enough data yet

Speed and success rate appear once this model has had at least 100 requests from several customers in the last 7 days.

What it costs in practice

Estimates at the current price. Coding agents send far more input than output; your own mix depends on the agent and the task.
WorkloadUSDBDT
One agent turn25K input, 1.5K output$0.04৳4.97
An hour of agent work1M input, 60K output$1.59৳198.75
A month of daily use20M input, 1.2M output$31.80৳3,975.00

Use DeepSeek V4 Pro (latest)

Each setup already points at Tokens with this model selected. Replace the key with one from your dashboard. Quickstart
~/.claude/settings.json
{
  "env": {
    "ANTHROPIC_BASE_URL": "https://tokens.bd",
    "ANTHROPIC_AUTH_TOKEN": "tok_live_your_key",
    "ANTHROPIC_MODEL": "deepseek/deepseek-v4-pro",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "deepseek/deepseek-v4-pro",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "deepseek/deepseek-v4-pro",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "deepseek/deepseek-v4-pro",
    "CLAUDE_CODE_SUBAGENT_MODEL": "deepseek/deepseek-v4-pro"
  }
}

Configure Claude Code CLI via ~/.claude/settings.json or environment exports.