xAI / Grok

The frontier challenger the synthesis credited with leading on context but hadn’t paged. xAI (Elon Musk, founded 2023) makes Grok; it sits with openai and anthropic (cross-wiki) on the proprietary frontier pole — but with a distinctive open-then-closed licensing arc and a very large context window.

The licensing arc — open, then closed

• Grok-1 (Nov 2023) was released under Apache-2.0 — genuinely open weights.
• Every version since (Grok-1.5 → 2 → 3 → 4 → 4.1) is proprietary (xAI-specific license). So Grok traces the opposite trajectory from the open-weight labs: it started open and closed up as it scaled — a useful counter-example to the “open weights only move one direction” read, and another point on the synthesis’s “open is a spectrum” thread.

Context & capability

• Context window grew from 128K (Grok-1.5) to 2M tokens in recent enterprise variants — the basis for the synthesis’s “xAI leads context” claim on llm-benchmarks.
• Grok-3 (Feb 2025) was trained with ~10× the compute of Grok-2 and claimed to beat GPT-4o on math benchmarks (AIME); Grok-4 (July 2025)/4.1 (Nov 2025) followed.
• Grok 4.5 (mid-2026) marks a strategy shift: xAI now competes on price, not the premium framing below. At $2/$6 per 1M tokens (grok-4-5-price-vs-benchmarks) it undercuts Opus 4.8 ($5/$25), GPT-5.5 ($5/$30), and Fable 5 ($10/$50). Its coding benchmarks trail (DeepSWE 53% vs Fable 5’s 70%), but xAI argues cost-per-task wins anyway — claiming 4.2× fewer tokens than Opus 4.8 and ~80 tok/s. The “good enough + cheap” deepseek/China playbook, now run by a US frontier lab. Musk’s own framing (per openai-gpt56-grok45-clash): an “Opus-class model, but faster, more token-efficient and lower cost,” launching “in the coming days” — timed hours after OpenAI’s GPT-5.6 rollout, the latest round of the Musk–Altman rivalry.
• Distribution moat: wired directly into X (Twitter), standalone iOS/Android apps, embedded in Tesla vehicles, plus an API — a distribution channel no other frontier lab has.

Where it sits

Grok started on the frontier-premium pole next to openai (proprietary, reasoning-focused, differentiated by context length and native X/Tesla distribution), but Grok 4.5 repositions it as a price leader (grok-4-5-price-vs-benchmarks) — the cheapest of the current frontier-adjacent models, betting that cost-per-task beats peak benchmarks. Its open→closed arc is the inverse of llama/qwen, making it a clean illustration that “open vs closed” is a strategy over time, not a fixed label — and now its pricing is on a similar trajectory. (Benchmark/context/price claims are vendor-stated, dated snapshots — the standing caveat.)

Related

openai · llm-benchmarks · llm-provider · open-weight-models · llama · qwen · grok-4-5-price-vs-benchmarks · openai-gpt56-grok45-clash · llm-api-pricing · deepseek