xAI

Executive Summary

What it is: xAI's coding stack is the open-source Grok Build CLI (Apache 2.0) plus an API now led by Grok 4.6 ($2/$6 per MTok under 200K prompt tokens, $4/$12 above, 500K context), with Grok 4.5 at the same headline rates and distribution in GitHub Copilot since August 14. Source: https://docs.x.ai/developers/models

What to watch out for: The July Grok Build exfiltration scandal (whole-repo uploads including .env secrets) still has no security advisory, deletion proof, or retention disclosure two months later; Grok 4.6 shipped without published benchmarks; and enterprise controls (SSO, audit, indemnity) are absent. Source: https://gist.github.com/cereblab/dc9a40bc26120f4540e4e09b75ffb547

Bottom line: Grok 4.6 at $2/$6 is frontier-adjacent coding at mid-tier prices and Musk's Grok 4.7 window lands in mid-September, but xAI remains the trust laggard of the tracked set; the CLI is still not safe for proprietary codebases. Source: https://docs.x.ai/developers/models

Key Terms

  • Grok 4.6 - xAI's new flagship (August; in GitHub Copilot August 14), recommended by docs for both code and chat. 500K context, knowledge cutoff February 1, 2026, $2/$6 per MTok under 200K prompt tokens ($4/$12 at or above), cached input $0.50/$1.00. Source: xAI – Models
  • Grok Build CLI - xAI's terminal coding agent (Apache 2.0, open-sourced July 15); remains the recommended harness for Grok coding workflows. Source: GitHub – Grok Build
  • Long-context pricing - requests whose prompt reaches 200K tokens are billed at the higher rate for all tokens in the request (not just the excess). Source: xAI – Models

Latest Changes

Changes since the 2026-07 report.

July watch items verified:

  • Grok Build exfiltration follow-up (security advisory, retention disclosure): still-pending. No security advisory, data-deletion confirmation, or retention timeline was found in August. The July incident (CLI uploading full git repos including .env files to a GCS bucket) remains unresolved from a disclosure standpoint. Source: Gist – Dc9A40Bc26120F4540E4E09B75Ffb547
  • Grok 4.7: still-pending (window is early-mid September per Musk). No model ID exists at docs.x.ai as of September 2. Source: Aitoolsrecap – Ai News August 31 2026

New August changes:

Plans

xAI is API-first; consumer Grok subscriptions exist but publish no coding-agent usage limits. Source: xAI – Models

API Pricing

Model Context Input $/MTok (<200K / >=200K) Cached input Output $/MTok (<200K / >=200K)
grok-4.6 500K $2.00 / $4.00 $0.50 / $1.00 $6.00 / $12.00
grok-4.5 500K $2.00 / $4.00 $0.30 / $0.60 $6.00 / $12.00
grok-4.3 1M $1.25 / $2.50 $0.20 / $0.40 $2.50 / $5.00
grok-4.20-0309 (reasoning/non-reasoning) 1M $1.25 / $2.50 $0.20 / $0.40 $2.50 / $5.00
grok-build-0.1 256K $1.00 / $2.00 $0.20 / $0.40 $2.00 / $4.00

Other: grok-imagine-image-2.0 $0.04/image; grok-imagine-video-1.5 $0.08/sec; voice $0.08/min (2.0). Source: xAI – Models

Model Performance / Benchmarks

No new xAI-published benchmark tables for Grok 4.6 were found on the models index; standing July figures for Grok 4.5: Artificial Analysis Intelligence Index 54, TerminalBench v2.1 83.3%, SWE-Bench Pro 64.7%. Source: xAI – Grok 4 5

Latest News

Grok 4.6 Ships Quietly (August 2026)

Grok 4.6 became the docs' recommended code and chat model in August and entered GitHub Copilot on August 14, without a flagship launch post comparable to Grok 4.5's July announcement. Grok 4.5 remains available at identical headline pricing. Sources: xAI – Models , GitHub – 2026 08 14 Grok 4 6 Is Now Available In Github Copilot

Grok 4.7 Window Approaches

Musk's stated window for Grok 4.7 is early-to-mid September; no model ID or docs page exists yet. Source: Aitoolsrecap – Ai News August 31 2026

Community Signals

No significant new HN threads on Grok 4.6 were found in August. The Grok Build exfiltration thread (539 points, July) remains the reference community discussion, with no follow-up disclosure from xAI. News – Item

Enterprise Readiness

Feature Available? Details
SSO (SAML/OIDC) No API keys only; not documented.
SCIM No Not documented.
Audit logs No Not documented.
IP indemnity No Not documented (unchanged).
Data residency No Not documented.
HIPAA No Not advertised.
Air-gapped / on-prem No Not offered.
SLA No Not published.
Admin controls (RBAC) No Not documented.
Batch API Partial Available for some models only. Source: xAI – Models

Transparency Gaps

Gap Details Severity
Exfiltration remediation No advisory, deletion proof, or CLI audit trail published two months after the Grok Build upload scandal. High
Grok 4.6 benchmarks No published scores accompany the new flagship's quiet rollout. Medium
Consumer coding limits No usage limits published for Grok subscriptions used with coding agents. Medium
Grok 4.7 timing "Early-mid September" per Musk, with no official date. Low