Model components / Fine-tuned / chat models

Grok 4.20

xAI Grok 4.20; beta Feb 17, 2026, API GA Mar 10, 2026 (Grok 4.20 + Multi-agent). Revision '4.20 0309 v2' dated Apr 7, 2026. Multi-agent architecture (4 parallel agents); 2M-token context.

xAI Grok 4.20; beta Feb 17, 2026, API GA Mar 10, 2026 (Grok 4.20 + Multi-agent). Revision '4.20 0309 v2' dated Apr 7, 2026. Multi-agent architecture (4 parallel agents); 2M-token context. Consolidated on 2026-07-29 from grok-3; openness follows grok-4-20, the release that currently governs. See sources/slug_aliases.yaml.

Openness

1 high confidence

1.0

weights: closed
data: closed
code: closed
license: Proprietary(API-only)

Proprietary, served through the xAI API, grok.com and X, with no downloadable weights, so 1/closed. Grok 4.20 governs and contributes capability 4; adoption 3 is carried from Grok 3, which had wider reach. The bare `grok` slug names this model line - the consumer chat app is a separate product recorded as grok-app.

https://docs.x.ai/developers/release-notes recorded 2026-07-28
xAI release notes record "Grok 4.20 and Grok 4.20 Multi-agent are live" and point at the API docs; no weight release accompanies the entry
https://docs.x.ai/docs/models recorded 2026-07-28
xAI's own model reference for the Grok 4.x line documents pricing, context limits and model-selection guidance only, and states no distribution or download path

Adoption

3 low confidence

3.0

Available to X users and via grok.com/xAI API in 2025 as the flagship Grok generation; now a superseded model behind Grok 4.x default. No clean standalone Grok-3 user/usage count published. Placed at 3 on broad X-surface availability as a former flagship, downgraded from app-surface 5 because xAI discloses no honest per-model figure and it is no longer the default. Directional.

https://x.ai/news/grok-3 recorded 2026-06-04
Grok 3 launched as xAI flagship reasoning model on X/grok.com

Capability

4 high confidence

4.0

Artificial Analysis Intelligence Index v4.0 score of 49, ranked #24 of 150 (median 36); index includes GPQA Diamond and SciCode (individual subscores not disclosed). Provisional LMSYS/LMArena Elo ~1505-1535 (vs 1483 for Grok 4.1). Strong but a notch below the GPT-5/Gemini 3.5/Opus 4.7 leaders on the AA index.

https://artificialanalysis.ai/models/grok-4-20 recorded 2026-06-04
flagship phase-C verification source
https://www.nextbigfuture.com/2026/02/xai-launches-grok-4-20-and-it-has-4-ai-agents-collaborating.html recorded 2026-06-04
flagship phase-C verification source

Unchanged since 2026-07-30 (last edited, not re-checked)