One family. Four models. Built for every scale.

Flagship reasoning for work where getting it wrong is expensive.

Nova is 4RGED’s highest-capability model. It is built for the hardest plans, the sharpest trade-offs, and multi-step agent runs where Magma is no longer enough.

What it does well

• Frontier-depth reasoning

  Handles ambiguous requirements, conflicting constraints, and multi-hop decisions without shortcutting to a shallow answer.

• Architecture & planning

  Breaks large changes into sequenced steps, surfaces risks early, and keeps system-wide context across long sessions.

• Multi-step agent execution

  Stays coherent across tool calls, retries, and verification loops — built for agents that must finish the job, not just draft a reply.

Choose this when

The hardest reasoning, architecture decisions, and critical agent workloads.

• The cost of a wrong architectural choice outweighs Nova’s higher unit price.
• You need the deepest multi-file or multi-service plan in the family — not a single-file patch.
• Your agents run the longest tool chains and must preserve intent across many steps.
• You are reviewing security-sensitive or regulated changes Magma still misses.

Customer use cases

• System design & architecture reviews

  Evaluate trade-offs across services, data stores, and rollout plans with explicit risks and alternatives.

• Long-horizon coding agents

  Drive agents that research, edit, test, and revise across a large codebase before proposing a merge-ready change.

• Hard refactors & migrations

  Plan and execute migrations where order-of-operations and compatibility matter as much as the code itself.

• Policy, security & correctness review

  Pressure-test PRs and agent output for subtle failures, unsafe patterns, and compliance gaps.

Deep reasoning for work where getting it wrong is expensive.

Magma sits just under Nova. It is built for long-horizon plans, hard trade-offs, and multi-step agent runs where depth matters more than raw speed — without Nova’s unit price on every request.

What it does well

• Frontier-depth reasoning

  Handles ambiguous requirements, conflicting constraints, and multi-hop decisions without shortcutting to a shallow answer.

• Architecture & planning

  Breaks large changes into sequenced steps, surfaces risks early, and keeps system-wide context across long sessions.

• Multi-step agent execution

  Stays coherent across tool calls, retries, and verification loops — built for agents that must finish the job, not just draft a reply.

Choose this when

Complex reasoning, architecture decisions, and critical agent workloads.

• The cost of a wrong architectural choice outweighs Magma’s higher unit price.
• You need deep, multi-file or multi-service plans — not a single-file patch.
• Your agents run long tool chains and must preserve intent across many steps.
• You are reviewing security-sensitive or regulated changes.

Customer use cases

• System design & architecture reviews

  Evaluate trade-offs across services, data stores, and rollout plans with explicit risks and alternatives.

• Long-horizon coding agents

  Drive agents that research, edit, test, and revise across a large codebase before proposing a merge-ready change.

• Hard refactors & migrations

  Plan and execute migrations where order-of-operations and compatibility matter as much as the code itself.

• Policy, security & correctness review

  Pressure-test PRs and agent output for subtle failures, unsafe patterns, and compliance gaps.

The production default — strong coding agents at a lower unit cost.

Blaze is the workhorse of the 4RGED family: excellent coding and agent quality with enough headroom for real production systems — without Magma pricing on every request.

What it does well

• Production coding quality

  Writes, edits, and explains code with strong repo awareness — tuned for IDE pair-programming and review loops.

• Reliable under load

  Built for everyday production traffic: consistent behavior across many sessions, orgs, and agent types.

• Agentic orchestration

  Coordinates tools, tests, and multi-agent workflows without needing flagship pricing on every turn.

Choose this when

Production workloads, coding agents, and day-to-day orchestration.

• You want one default model for IDE, Cloud, and CI agents.
• Coding quality matters, but Magma-level depth is not required on every call.
• You run many parallel agent sessions and need predictable cost.
• You need a million-token context window for real repositories and tickets.

Customer use cases

• IDE pair-programming agents

  Inline edits, explanations, and multi-file changes while an engineer stays in flow.

• Cloud parallel agent fleets

  Spin up many agents on tickets, PRs, or migrations without flagship cost on each run.

• CI, CLI & automation

  Fix failing checks, generate patches, and drive scripted workflows from the terminal or pipelines.

• Team default for new projects

  Standardize on one model so policy, prompts, and evals stay simple across the org.

Maximum context and throughput for everyday high-volume work.

Ember is optimized for scale: the largest context window in the family at the lowest price — ideal when you need to read a lot, respond quickly, and run many jobs.

What it does well

• High-throughput responses

  Built for queues and interactive chat where latency and cost per request dominate.

• Large document & dataset intake

  Ingest long docs, transcripts, and dumps in fewer passes — less glue code, fewer missed sections.

• Efficient long-context processing

  2M input and output context for jobs that would otherwise require aggressive truncation.

Choose this when

Everyday chat, summaries, triage, and high-volume long-context jobs.

• You process large documents or long conversations in a single request.
• Unit economics matter more than maximum reasoning depth.
• You need the fastest turnaround in the 4RGED family.
• You want a cheap first-pass model before escalating to Blaze, Magma, or Nova.

Customer use cases

• Customer & internal chat assistants

  Fast answers across long histories without burning Magma or Blaze budget on every message.

• Summaries & briefings

  Collapse tickets, meetings, and docs into actionable notes at high volume.

• Classification & triage pipelines

  Label, route, and prioritize incoming work before a deeper model gets involved.

• Large-context ingestion

  Pull big files and datasets into agents for search, extract, and transform workflows.