Mirendil Fern (10M Context)
Purpose-built 10-million-token context architecture capable of ingesting entire monorepos, multi-gigabyte kernel trees, and weeks of production telemetry in a single unchunked prompt.
Interactive Token Cost Estimator (Mirendil Fern 10M)
Architectural Strengths & Target Workloads
- True 10,000,000-token context window with 99.2% multi-needle retrieval accuracy at 9.5M depth
- Hierarchical Sparse Attention slashes KV-cache memory by 78% compared to dense ring attention
- Industry-lowest $0.18 / 1M cached token cost for persistent codebase indexing
Primary Engineering Workloads: Whole-Monorepo Ingestion · Kernel & Firmware Auditing · Cross-Repository Migration.
Frequently Asked Questions (Mirendil Fern 10M)
How much does Mirendil Fern (10M Context) cost per 1M tokens?
Mirendil Fern (10M Context) is priced at $1.40 per 1M uncached input tokens, $0.18 per 1M cached input tokens, and $5.60 per 1M output tokens (Mirendil API / Enterprise VPC (78% lower KV-cache footprint via HSA)).
What is the context window of Mirendil Fern (10M Context)?
It supports up to 10,000,000 tokens (10,000,000 tokens) and a maximum generation output of 131,072 tokens.
OpenAI Ships GPT-6.1 Terra at Half the Price, While Newcomer Mirendil Debuts Fern with a 10M-Token Context
A cheaper, faster computer-use model lands in ChatGPT Work, Codex, the API and Amazon Bedrock on the same day a new lab bets on extreme context length for agent workloads.
Read Full Technical Dispatch (5 min read) →