Model
Model explorer
Grok 4.20 Multi-Agent
CLOSEDSpaceXAI · Grok 4.20 family · released Mar 10, 2026
Multi-agent sibling of Grok 4.20, paralleling how 'Grok 4 Heavy' relates to 'Grok-4' elsewhere in this corpus: 4 specialized agent roles collaborate at inference time on every complex query, running as parallel replicas of the same underlying model rather than 4 independent models. xAI states the marginal inference cost is roughly 1.5-2.5x a single pass, not 4x. Public beta Feb 17, 2026; full API GA Mar 10, 2026 alongside the base model.
ReasoningCodingVisionFunction callingTool useAgentic
—
Elo · unrated
Parameters
Undisclosed
Active params
Undisclosed
Context
1M tokens
Architecture
Mixture-of-Experts (size undisclosed) served as 4 parallel specialized agent roles (coordinator 'Grok/Captain', research/fact-check 'Harper', math/logic 'Benjamin', creative/balance 'Lucas') sharing weights and KV-cache, debating before synthesizing one final answer
License
Proprietary
Languages
—
API price (in/out)
$1.25 / $2.5
Modalities
text · vision
Benchmark results
Bar shows position within the tracked field; marker = field best
Run it locally
Closed weights — available via API only. No local deployment.
Input / M tok
$1.25
Output / M tok
$2.5
Grok 4.20 family
Elo progression across releases
API price $1.25/$2.5 · each benchmark row carries its own source badge (see methodology)