Grok 4.6 matches frontier rivals at half the price

xAI priced Grok 4.6 at $2 per million input tokens and $6 per million output tokens, roughly half the sticker price of comparable frontier models from OpenAI and Anthropic.

xAI's Grok 4.6 matches OpenAI's GPT-5.6 Sol and Anthropic's Fable 5 on frontier benchmarks while pricing tokens at roughly half the rate, a cost move aimed at developers whose bills scale with usage.

"It delivers frontier intelligence and is a significant improvement over Grok 4.5 at the same price," xAI said in its Aug. 12 launch announcement, with Chief Executive Elon Musk calling the release "a banger" on X.

On the Artificial Analysis Intelligence Index, a composite of nine benchmarks, Grok 4.6 scored 61, tying GPT-5.6 Sol Max and trailing Fable 5 Max's 62. Grok 4.5 High scored 56. The 1.5-trillion-parameter model, built on the same V9 foundation as Grok 4.5, led on GDPVal-AA v2 (1753 vs 1741 for Fable 5 Max) and AA-Briefcase (1577 vs 1574), while trailing on DeepSWE v1.1 (65.9% vs 73% for GPT-5.6 Sol Max) and Terminal-Bench v3.0 (26% vs 34.6%).

The pricing is the sharper competitive weapon. At $2 per million input tokens and $6 per million output tokens, Grok 4.6 undercuts GPT-5.6 Sol Max and Fable 5 Max by roughly half, and xAI is doubling included usage inside Cursor and Grok Build for the first week to drive adoption among developers.

The pricing story is the real headline

xAI's messaging is explicit: "half the price of other frontier models." That framing targets the developer and agent-builder cohort whose token bills scale with volume — the exact segment xAI needs to win to close the API market-share gap with OpenAI and Anthropic. Pairing the price with a first-week 2x usage promotion inside Cursor, one of the most popular AI-native code editors, is a distribution play: Cursor users will test Grok 4.6 against their default model at no marginal cost.

The model is live today across the xAI API, Grok Build, Cursor, and Grok Bot on X, with third-party access through OpenRouter, Vercel, and Cloudflare. A faster variant is available at double the base rate for lower-latency workloads.

Benchmarks land in the same tier

Across the full eval set, Grok 4.6 beats Grok 4.5 High on every benchmark by a wide margin and is genuinely competitive with GPT-5.6 Sol Max and Fable 5 Max — leading on a few evals, trailing on others, but never falling far behind. Musk cited a 1753 ELO score, though that figure is xAI-reported and has not yet been independently verified on public arenas like LMSYS Chatbot Arena.

xAI says the model went through a longer secondary training phase than its predecessor, layering in curated, model-generated data aimed at reasoning and technical depth, along with engineering-heavy datasets. The supervised fine-tuning stage used Grok 4.5 to regenerate training trajectories across reasoning efforts, agent setups, and domains including STEM and software engineering, filtering out low-quality examples with automated checks.

What it means for the AI economy

The release cadence is aggressive. Grok 4.7, a larger 2.1-trillion-parameter model, is expected within weeks, and Grok 5 — which will incorporate the full corpus of SpaceX engineering data — is targeted before the end of 2026. SpaceX, which merged with xAI and is acquiring developer-tools company Cursor for $60 billion in stock, reported $2.56 billion in AI segment revenue in the second quarter, up 247 percent from a year earlier. Musk has said AI revenue will exceed all other SpaceX units by September, with a potential $300 billion to $500 billion annual run rate tied to 10 gigawatts of AI computing capacity by the end of next year.

For Tesla, which is wiring Grok into vehicles for in-car voice interaction and, longer term, into Optimus, a cheaper, faster frontier model improves the economics of shipping AI features across millions of endpoints. Every cent of token cost is multiplied by fleet scale.

The next 72 hours of independent benchmark data will decide whether Grok 4.6 is genuinely competitive at the top of the frontier stack, or whether the ELO claim outpaces reality. Either way, the pricing move is real and puts direct cost pressure on every other frontier lab's API business.

This article is for informational purposes only and does not constitute investment advice.