Command A+

Released
2026-05-20
Last refreshed
2026-10-17
Status
Researched 33d ago
Open sourceCommercial use: permittedMultimodalRAGAgentsLong contextVisionJSON / Tool use
Cohere releases · 10 in the last 12 months · this family litChangelog →
Specifications
Family
Command
Released
2026-05-20
Context
128K
Max output
64K
Parameters
218B total / 25B active
Architecture
Mixture of Experts
Knowledge cutoff
2025-04
Specialization
general
Openness
Open source
License
Apache 2.0OSI-approvedPermitted
Weights
Available
Code
Unknown
Training
Pretrained
Created by

Empowering developers with advanced language AI.

Toronto, Ontario, Canada
Founded 2022
Website
Pricing
Output / 1M
$1.50
Input / 1M
$0.300

Cheapest of 3 routes · OpenRouter · cache read $0.150

About

Command A+ is Cohere's open-weight sparse mixture-of-experts model for enterprise agentic, multimodal, multilingual, RAG, and reasoning-heavy workloads. It combines text and image inputs, tool use, structured outputs, 48-language support, and hardware-efficient deployment that can run on one B200 or two H100 GPUs.

Providers(3)

Compare all 3

Compare API pricing across 3 tracked provider routes for input and output tokens, batch, and cached reads when available.

ProviderInput / 1MOutput / 1MCacheRoute
OpenRouter$0.300$1.50read $0.150
Serverless
Microsoft Foundry$0.800$3.20-
Serverless
Cohere APIContact salesContact sales-
ServerlessProvisioned

Available via routers & gateways(6)

Capabilities

VisionMultimodalReasoningJSON / Tool useStructured Outputs

Benchmark peer barsfor Vision

Benchmark scores(3)

Scores are benchmark-specific and are direction-aware: the same numeric gap can mean very different outcomes across suites. Use the leaderboard context and this model's provider route to decide whether the winning margin is meaningful for your workload.
BenchmarkScoreVersionEvaluationSource
90.0AIME 2025 (accuracy)Observed 2026-06-07—
75.1MMMU (accuracy)Observed 2026-06-07—
26.0WMT26 General MT findings, Figure 3 Avg. Rank (lower is better)Observed 2026-10-27
Configuration: Command A+
Harness: Contrastive Error Span Annotation (cESA), human annotators
Evaluator: WMT26 organizers
Confidence: confirmed
Notes: WMT26 General MT findings (2026.wmt-1.48), Figure 3: Avg. Rank 26.0 for the system listed as 'Command A+', averaged over the 23 language pairs it entered; rank 1 is best, so lower is better (not a 0-100 score). Per-pair cESA scores (0-100, higher is better, not comparable across pairs): cs-vi 73.2, cs-de 71.0, en-cs 61.3, en-de 70.8, en-is 42.0, en-zh-Hant 61.9, en-arz 80.1, en-ru 70.4, en-be 53.7, en-lld 2.9, en-se 2.6, en-et 56.6, en-ja 81.4, en-kk 57.8, en-id 76.8, zh-ja 66.9, en-ko 63.0, en-th 70.1, en-lij 7.0, cs-uk 76.9, en-uk 75.9, en-hy 12.5, en-zh-Hans 65.8. Organizer baseline 'Command A+', not the Cohere CAT+ submission. Source: WMT26 organizers. Retrieved 2026-10-27.
Source

Source: Findings of the WMT26 General Machine Translation Shared Task (Kocmi et al., 2026)

Migration checks

No linked migration route is available for this model yet.

Compare Command A+ with other models