Nex-AGI

Nex-N2.5-Max

Name: Nex-N2.5-Max
Author: Nex-AGI

Nex-N2.5-Max is a Nex-AGI model with 1.05M context, Not verified in / Not verified out per million tokens, last verified October 2, 2026. Open weights: yes. Public-board scores are linked to the publisher; vendor-only figures stay in the watch-out, not on the board.

A downloadable text-only MoE model for agent workflows, with a publisher-provided multi-node serving example.

Recorded profile verification October 2, 2026Individual provider fields may change

AI model specification details
Specification	Nex-N2.5-Max
Provider	Nex-AGI
Tier	Frontier
Context window	1.05M
Max output	Not verifiedUnverified
Input / 1M tokens	Not verifiedUnverified
Output / 1M tokens	Not verifiedUnverified
Weights	Open
Parameters	1.6T total MoE; active parameter count not verified
Reasoning levels	none, medium, high
Modalities	text
License	Apache-2.0 (publisher repository metadata)
Released	Not verifiedUnverified

Pricing tiers: No verified hosted API tariff. Downloadable weights incur self-host compute costs. The checkpoint configuration has a 1,048,576 position limit, while the publisher deployment example sets serving context to 262,144; no maximum output cap is verified.

Publisher evidence

Published benchmark results

Each result keeps its source date visible. Dates refer to source publication or capture unless the source explicitly states when it measured the result. A missing benchmark is not treated as a zero.

Published benchmark results for Nex-N2.5-Max
Benchmark	Score	Source date	Source
Terminal-Bench 2.1Vendor NexAU coding harness; temperature 0.7, top-p 0.95, top-k 40; reasoning effort and experiment date unpublishedCross-model comparability not verified	86.1%	2026-10-01	Nex-AGI Nex-N2.5 publisher evaluation table · View source
DeepSWE 1.1Vendor NexAU coding harness; temperature 0.7, top-p 0.95, top-k 40; reasoning effort and experiment date unpublishedCross-model comparability not verified	65.6%	2026-10-01	Nex-AGI Nex-N2.5 publisher evaluation table · View source

Best for

Evaluating self-hosted agent workflows

Watch out

Release day is unverified; repository creation and route version strings are not release receipts. Apache-2.0 is declared in publisher metadata; a standalone licence file remains unverified. reasoning_effort=medium is adaptive and the default; none disables reasoning and high always enables it. The chat template directly consumes reasoning_effort; enable_thinking/thinking_mode require gateway translation. Use qwen3 reasoning parser for Pro/mini, deepseek-r1 for Max, and qwen3_coder for tools. Vendor NexAU scores have incomplete run/effort metadata, so verdicts are withheld. Native config gives 1,048,576 positions, while the example serves only 262,144. This 1.6T checkpoint requires substantial multi-node hardware.

Source receipts

Catalog figures for this model were checked against the following sources.

Publisher family and exact template effort contract (accessed 2026-10-02)
Exact downloadable checkpoint and declared licence (accessed 2026-10-02)
Exact native position config and quantization (accessed 2026-10-02)

Common questions

Nex-N2.5-Max

Answered from the verified figures on this page rather than general guidance.

What is Nex-N2.5-Max's context window?

Nex-N2.5-Max accepts about 1.05M tokens of context. That only matters if you routinely send very long documents, large codebases, or multi-turn histories that approach that limit.

What is Nex-N2.5-Max best for?

Nex-N2.5-Max is a frontier tier from Nex-AGI. It suits evaluating self-hosted agent workflows. Release day is unverified; repository creation and route version strings are not release receipts. Apache-2.0 is declared in publisher metadata; a standalone licence file remains unverified. reasoning_effort=medium is adaptive and the default; none disables reasoning and high always enables it. The chat template directly consumes reasoning_effort; enable_thinking/thinking_mode require gateway translation. Use qwen3 reasoning parser for Pro/mini, deepseek-r1 for Max, and qwen3_coder for tools. Vendor NexAU scores have incomplete run/effort metadata, so verdicts are withheld. Native config gives 1,048,576 positions, while the example serves only 262,144. This 1.6T checkpoint requires substantial multi-node hardware.

Can I self-host Nex-N2.5-Max?

Nex-N2.5-Max publishes open weights, but self-hosting depends on the licence, hardware footprint, quantisation quality, and serving stack. A hosted API is often cheaper until you have measured throughput and concurrency on your own hardware.

Model comparison catalog Benchmark library