SpaceXAI Grok 4.6 ties OpenAI's best model and undercuts on price
SpaceXAI's Grok 4.6 has reached parity with OpenAI's flagship GPT-5.6 Sol on the Artificial Analysis Intelligence Index, both scoring 61, while undercutting rivals on price at $2 per million input and $6 per million output tokens—more than 60% cheaper than Claude Opus 5 and GPT-5.6 Sol. Crucially, it excels at agentic tasks, completing complex workflows in about 53 steps versus Claude's 103, signaling that cost-efficiency and agent-specific performance are becoming decisive differentiators. The model, trained and running on NVIDIA's GB300 NVL72, is available via API and Cursor, with double usage quotas for the first week. However, benchmark claims remain vendor-reported and third-party verification is pending, so builders should validate performance against their own workloads.
SpaceXAI Grok 4.6 ties OpenAI's best model and undercuts on price
According to the Artificial Analysis Intelligence Index, SpaceXAI's new model scores 61 points, tying OpenAI's GPT-5.6 Sol. Pricing stays at $2/$6 per million tokens, over 60% cheaper than Claude Opus 5 and GPT-5.6 Sol.
Key takeaway
Agentic efficiency and price are becoming more valuable differentiators than raw intelligence scores in the current AI landscape.
What happened
According to The Decoder, SpaceXAI's new Grok 4.6 model scores 61 points on the Artificial Analysis Intelligence Index, matching OpenAI's GPT-5.6 Sol and trailing only Anthropic's Claude Opus 5 at 63. The model shows a five-point jump over its predecessor Grok 4.5.
On agentic tasks, it completes complex workflows in about 53 steps, whereas Claude Opus 5 needs roughly 103, and pricing remains at $2 per million input and $6 per million output tokens, over 60% cheaper than Claude Opus 5 and GPT-5.6 Sol. NVIDIA also announced that Grok 4.6 runs and is trained on the GB300 NVL72 system.
Evidence
Grok 4.6 scores 61 on the Artificial Analysis Intelligence Index, tying OpenAI's GPT-5.6 Sol.
The Decoder · attributed
xAI's Grok 4.6 scores 61 points on the Artificial Analysis Intelligence Index, tying GPT-5.6 Sol and trailing only Anthropic's Claude Opus 5.
On agentic tasks, Grok 4.6 completes complex workflows in about 53 steps, while Claude Opus 5 needs 103.
The Decoder · attributed
On agentic tasks, it completes complex workflows in about 53 steps where Claude Opus 5 needs 103.
Pricing is $2 per million input and $6 per million output tokens, over 60% cheaper than Claude Opus 5 and GPT-5.6 Sol.
The Decoder · attributed
Pricing stays at $2/$6 per million tokens. That's more than 60 percent cheaper than Claude Opus 5 ($5/$25) and GPT-5.6 Sol ($5/$30).
Grok 4.6 is trained and runs on NVIDIA's GB300 NVL72.
NVIDIA (X) · attributed
Congrats to the @SpaceXAI team on the release of Grok 4.6. Grok 4.6 brings frontier intelligence, running and trained on NVIDIA GB300 NVL72 with NVLink to deliver exceptional performance, reliability and lowest token cost.
Why it matters
Builders and operators can significantly reduce inference costs and latency for agentic workflows by adopting Grok 4.6, making it a strategic choice for high-volume, multi-step automation tasks.
Limits and uncertainties
Details on the Grok Bot product from The Information were inaccessible due to browser context closure, leaving product specifics incomplete.
The reported Gemini 3.5 Pro release and OpenAI usage reset pricing are based on unofficial reports and unconfirmed.
Practical implications
Teams should evaluate Grok 4.6's agentic performance on their own workloads, given its 60% cost advantage, and consider migrating high-volume automation to reduce operational expenses.
The availability on Cursor and partners like OpenRouter means integration is straightforward, but validate quota doubling terms before committing.
What to watch
Watch for independent verification of Grok 4.6's benchmark scores and real-world agentic performance.
Monitor competitive price responses from OpenAI, Anthropic, and Google.
