Model updates & releases

What's new with NLC AI models — release notes, deep dives, and usage tips.

Meet the models

FLAGSHIP
NLC ULTRA MAX 1.0

NLC ULTRA MAX 1.0

Flagship multi-agent

The flagship multi-agent powerhouse. A router classifies your task, then dispatches FRONTEND work to a vision-capable executor and BACKEND work to a deep-reasoning executor. 1M context. Best for full-stack features and the hardest reasoning + visual tasks in one call.

Context1,048,576
Max output1,044,000
Input price$3.50/M
Output price$11.00/M

Best for:

Full-stack features Frontend + Backend Multi-agent Hardest tasks
NLC Vision 1.0

NLC Vision 1.0

Multimodal

Full multimodal model — understand images and code in one call. 256K context. Best for visual tasks, screenshots, diagrams, UI analysis.

Context262,144
Max output200,000
Input price$1.80/M
Output price$5.50/M

Best for:

Image understanding Screenshots Diagrams UI analysis
NLC Fast 1.0

NLC Fast 1.0

Cheapest

The cheapest, fastest model. 1M context. Great for high-throughput coding, agentic workloads, and quick answers. Best price-to-speed ratio.

Context1,048,576
Max output900,000
Input price$0.45/M
Output price$0.90/M

Best for:

Quick answers High-throughput Agentic workloads CI/CD
PRO
NLC PRO 1.0

NLC PRO 1.0

Deep reasoning

Single-model deep reasoning + code. 1M token context, ~1M output, vision-capable. Best when you want the strongest single executor without the router overhead — code analysis, long documents, large codebases.

Context1,048,576
Max output1,044,000
Input price$3.20/M
Output price$9.80/M

Best for:

Code analysis Deep reasoning Long documents Large codebases
BUDGET
NLC ShipLow 1.0

NLC ShipLow 1.0

Budget fusion

Six-lane budget fusion. The router reads your task and dispatches to the specialist executor that's strongest at it — code, agentic code, math, vision, multilingual, or general chat. Vision + MCP tools at the lowest price tier.

Context131,072
Max output65,536
Input price$0.30/M
Output price$0.60/M

Six lanes:

General Code Agent Math Vision Multilingual
NLC Embed 1.0

NLC Embed 1.0

RAG

Embedding model for RAG and semantic search. Variable-length output. Best for building retrieval-augmented generation pipelines.

Context8,192
Input price$0.20/M

Best for:

RAG Semantic search Document similarity
NLC Rerank 1.0

NLC Rerank 1.0

RAG

Rerank documents against a query for better retrieval. Use as a second stage after vector search to boost RAG quality.

Context32,768
Input price$0.20/M

Best for:

Second-stage RAG Document reranking Retrieval boost

Release notes

NEW · BUDGET v2.1.0 · Jun 24, 2026

NLC ShipLow 1.0 is live — six-lane budget fusion for everything-at-once

ShipLow is a budget multi-agent that doesn't compromise on capabilities. A fast router reads your request and dispatches it to the specialist executor that's strongest at that specific kind of work — six dedicated lanes — all under one nlc-shiplow brand id. $0.30/M input, $0.60/M output — a fraction of NLC PRO.

The six lanes

  • · General chat & reasoning — fast all-rounder lane with tool-use and image input.
  • · Code — single-file generation, refactors, and bug fixes at competition-grade quality.
  • · Agent — multi-file and repo-level engineering with patch-style edits.
  • · Math — careful step-by-step numeric reasoning and proofs.
  • · Vision — image attached → auto-routed here. Reads screenshots, diagrams, UI mockups.
  • · Multilingual — Thai / Indic / CJK / Arabic, natural and fluent.
  • Auto-routing — the router classifies in <1s and dispatches. You see one streamed answer.
  • 128K context — fits a multi-file repo, a long doc, or an hours-long conversation.
  • MCP tools work — call your connectors from any lane.
  • $0.30 / $0.60 per 1M tokens — over 10× cheaper than NLC PRO and 15× cheaper than NLC ULTRA MAX. Drop back to those any time you want maximum quality.
Budget Multi-agent Vision MCP tools 128K context
NEW v2.0.0 · Jun 20, 2026

NLC ULTRA MAX 1.0 is live — multi-agent routing for full-stack work

ULTRA MAX is our premium tier: a fast router classifies your request as FRONTEND, BACKEND, or GENERAL, then dispatches it to the best executor model. Frontend tasks go to a vision-capable model; backend tasks go to a deep-reasoning model. You get one streamed response under the nlc-ultra brand id — you never see which sub-agent ran.

Multi-agent Full-stack 1M context
UPDATE v1.9.0 · Jun 20, 2026

Migrated to a new inference provider — faster, more reliable

We've switched our upstream inference to a new provider for better latency and reliability. All models (PRO, Vision, Fast, ULTRA MAX, Embed, Rerank) now run on the new infrastructure. No API changes — your existing keys and code work as-is.

Infrastructure Performance
FEATURE v1.8.0 · Jun 20, 2026

Anthropic Messages API now supported

You can now use the Anthropic Python/JS SDK with NLC AI. Point your client at https://oddsforge.org/inference and use your NLC API key. Works with Claude Code, Anthropic SDK, and any Anthropic-compatible client.

Anthropic Claude Code New protocol
FEATURE v1.7.0 · Jun 19, 2026

NLC Embed 1.0 + NLC Rerank 1.0 — RAG pipeline endpoints

Two new endpoints for building RAG pipelines: /v1/embeddings generates vector embeddings, and /v1/rerank reranks documents against a query. Both at $0.20/M input tokens.

Embeddings Rerank RAG
SECURITY v1.6.0 · Jun 19, 2026

MFA, API key scopes, audit logs, GDPR export

Major security update: TOTP two-factor authentication, per-API-key scopes (e.g. embeddings-only), admin + user audit logs, webhook delivery logs with retry, SSRF guards, GDPR data export and account deletion.

MFA Scopes GDPR Audit log
Start using NLC AI — free