Command R+ is the upgraded Command R from August 2024, and inside Atlas its selling point is tool-calling reliability. Command R+ keeps the 128,000 token context and 4,000 token output of Command R but is priced at Cohere's flagship rate of $2.5 in and $10 out per Mtok, buying materially better tool-calling and RAG reliability. That matters in Atlas because every Atlas action, whether read, edit, grep, or bash, is a tool call, so a model that picks the right tool and fills its arguments correctly saves more turns than it costs.
Why does Command R+ tool calling matter in Atlas?
Command R+ has stronger multi step tool use than Command R, and that matters because every Atlas action, whether read, edit, grep, or bash, is a tool call. A model that misfires on tool arguments burns turns. Command R+ shipped in August 2024 specifically to fix that.
Atlas is a tool-call machine. Atlas searches code with hybrid semantic and keyword retrieval fused by reciprocal rank fusion, Atlas computes a unified diff for every file edit and surfaces it for approval before writing, and Atlas connects to Model Context Protocol servers and exposes their tools to the agent. Each of those is a schema the model must fill correctly. Command R+ was built with multi step tool use as a first-class capability, which is why it is worth the $2.5 per Mtok input over Command R's $0.15 in the main loop even though both models see the same 128,000 tokens. The value is not a bigger window. The value is fewer wasted turns inside the window you already have.
How does Command R+ handle Atlas's code index?
Command R+ pairs a 128,000 token context with citation-aware RAG behavior, a good fit for Atlas's tree-sitter AST chunked code index. Atlas indexes code by AST declarations using tree-sitter, not blind line windows, so every chunk Command R+ receives is a whole declaration it can cite.
Citation-aware RAG behavior means Command R+ tracks which retrieved chunk supports which claim, and Atlas hands it chunks worth citing. Because Atlas indexes code by AST declarations using tree-sitter rather than blind line windows, a retrieved chunk is a whole function or a whole type, with its name and signature intact. Command R+ can then say, in effect, this change belongs in this declaration, and be right about where. Feed a model ragged line windows and citation awareness degrades into guessing. The 128,000 token context is enough to hold a meaningful set of those AST chunks alongside the diff under discussion, which is the working set most single-feature tasks actually need.
What is the catch with Command R+ pricing?
The catch with Command R+ is the 4,000 token output ceiling, the smallest of any model at this $2.5 and $10 per Mtok price point. Command R+ charges Cohere's flagship rate while emitting less code per turn than much cheaper models from other vendors.
Command R+ is priced at $2.5 per Mtok input and $10 per Mtok output, the same as Command A, but Command A gives you 256,000 tokens of context and 8,000 tokens of output for that money while Command R+ gives you 128,000 and 4,000. Stated plainly: Command A offers 2x the context for the same price, and Command R+ is now superseded by it. Atlas's own documented guidance is blunt about the consequence. If you hit the 4,000 token output wall often, move up to command-a-03-2025. Command R+ remains a defensible choice when tool-calling reliability specifically is what you are buying, but nobody should choose Command R+ for its price-to-context ratio in 2026.
How should you configure permissions for Command R+ in Atlas?
Set permission mode to ask on write tools so you approve each of Command R+'s small patches. Because Command R+ caps output at 4,000 tokens, a real change arrives as a sequence of small diffs, and Atlas gates every one of them before it touches disk.
Every Atlas tool call is permission-gated against allow, ask, and deny rules before it runs, and Command R+'s output shape makes ask the right setting on writes. A 4,000 token ceiling turns one logical change into several patches, and reviewing each as it lands is cheaper than untangling a half-applied refactor later. Atlas computes a unified diff for every file edit and surfaces it for approval before writing, so the ask gate costs you a keystroke, not a context switch. Atlas also snapshots file changes as git patches so edits can be diffed and rolled back, which means even an approved patch is reversible. For Command R+ specifically, the combination of small patches plus ask-on-write plus git patch snapshots is the configuration that keeps a non-reasoning model safe to run on real code.
When should you pick a different model than Command R+?
Pick Command A over Command R+ when you want more room. Command A offers 2x the context at 256,000 tokens for the same $2.5 and $10 per Mtok price, plus an 8,000 token output instead of 4,000. Command R+ is non-reasoning and superseded by Command A.
Three concrete triggers to move off Command R+. First, if you hit the 4,000 token output wall often, move up to command-a-03-2025, which doubles the generation budget. Second, if you need actual reasoning, Command R+ is non-reasoning and leans on Atlas's read-only plan agent to do the decomposition, so a reasoning checkpoint would carry more of that load itself. Third, if the work is cheap background traffic, Command R+ at $2.5 per Mtok input is a waste and Command R at $0.15 does the same job. Atlas lets you switch the active model and provider on the fly with favorites and recents, so pinning both Command R+ and command-a-03-2025 and flipping between them mid-task takes one keystroke.
Setup
- 01Export COHERE_API_KEY from the Cohere dashboard.
- 02Run `atlas models cohere` and locate command-r-plus-08-2024.
- 03Pin "model": "cohere/command-r-plus-08-2024" in atlas.json.
- 04Set permission mode to ask on write tools so you approve each of its small patches.
- 05If you hit the 4,000 token output wall often, move up to command-a-03-2025.
Frequently asked questions
- how much does Command R+ cost per million tokens
- Command R+ costs $2.5 per Mtok input and $10 per Mtok output, Cohere's flagship rate. That is the same price as Command A, which offers 2x the context.
- what is Command R+ context window
- Command R+ has a 128,000 token context window and a 4,000 token output ceiling, the same as Command R. Command A doubles the context to 256,000 tokens for the same price.
- Command R+ vs Command A for a coding agent
- Command R+ and Command A both cost $2.5 per Mtok input and $10 per Mtok output, but Command A gives 256,000 tokens of context and 8,000 of output versus 128,000 and 4,000.
- how do I configure Command R+ in Atlas
- Export COHERE_API_KEY, run `atlas models cohere` to locate command-r-plus-08-2024, then pin "model": "cohere/command-r-plus-08-2024" in atlas.json.
- is Command R+ good at tool calling
- Yes. Command R+ has stronger multi step tool use than Command R, which matters because every Atlas action, whether read, edit, grep, or bash, is a tool call.
- can Command R+ run in a private VPC
- Yes. Command R+ shares the same private VPC and on-prem deployment story as the rest of the Command family, so Atlas can be pointed at an internal endpoint.
- why does my Command R+ output keep getting cut off
- Command R+ caps output at 4,000 tokens, the smallest of any model at the $2.5 and $10 per Mtok price point. If you hit that wall often, move up to command-a-03-2025.
Try Atlas in your terminal
The terminal-native AI coding agent. Free core, single binary.
Install AtlasRelated guides
Atlas vs Poolside: Terminal AI Coding Agents in 2026
Compare Atlas and Poolside, two terminal AI coding agents in 2026. Atlas offers a free core and robust safety features, while Poolside provides open-weight Laguna models and on-prem deployment options.
Atlas for F#: A Terminal-Native AI Coding Agent for .fsproj Solutions in 2026
Atlas is a terminal-native AI coding agent for F# in 2026. It respects .fsproj file order, maps discriminated unions, runs dotnet test behind a prompt, and runs Fantomas.
Atlas for Actix Web in 2026
Atlas is a terminal-native AI coding agent for Actix Web in 2026. It reads extractors and app_data, then runs cargo test and cargo clippy behind a prompt.
Atlas vs Ellipsis: Terminal AI Coding Agents in 2026
Compare Atlas, a terminal-native AI coding agent with free core and local embeddings, against Ellipsis, a cloud platform with usage-based pricing and live session tracing for 2026.
Atlas for Clojure: A Terminal-Native AI Coding Agent for deps.edn and Kaocha in 2026
Atlas is a terminal-native AI coding agent for Clojure in 2026. It reads deps.edn aliases and namespace requires, runs clojure -M:test with Kaocha, and applies cljfmt.
Atlas vs Cline: Terminal AI Coding Agents in 2026
Compare Atlas and Cline, two leading AI coding agents for 2026. Atlas offers terminal-native TUI and permission-gated tools, while Cline integrates with VS Code and provides checkpoint rollbacks.
Atlas for Godot: Terminal-Native AI Coding for GDScript and the Node Tree in 2026
Atlas is a terminal-native AI coding agent for Godot in 2026, working across GDScript, the node tree, and signals, where scene structure is half the program.
Atlas for TypeScript in 2026
In 2026, TypeScript developers leverage Atlas, the terminal-native AI coding agent, to enhance productivity. Atlas understands your types, ensures code quality, and offers robust safety features.