New models from Meta and SpaceX scored near the leaders on a widely watched independent benchmark. Axios reported that SpaceX's Grok 4.6 scored roughly even with OpenAI's GPT-5.6 Sol Max and just behind Anthropic's Fable 5 Max on the Artificial Analysis index. Meta and SpaceX released new AI models this week. Both companies improved their standing in frontier-model comparisons. The checked account from Axios establishes those points through direct reporting, public records or attributed statements; where an institution or participant made a claim, that claim remains attributed rather than converted into an independently verified conclusion.
Grok 4.6 scored approximately even with OpenAI's GPT-5.6 Sol Max on the cited index. It scored just behind Anthropic's Fable 5 Max. Elon Musk said Grok 4.7 could arrive within several weeks. The ranking came from the Artificial Analysis Intelligence Index. These details define what changed by the edition deadline. They also separate the event itself from forecasts about what it may produce. Dates, quantities and official actions are stated only to the precision supported by the source record, and later updates could refine the sequence without erasing the confirmed core.
Benchmarks are useful comparative measurements but can be optimized for and revised. Model cost, latency, tool use and safety can matter as much as one composite score. Company predictions about future releases are not confirmed delivery dates. In practical terms, that context explains the institutions, infrastructure and constraints surrounding meta and spacex narrow the gap in the frontier-model race. It does not settle disputed motives or predict outcomes. Instead, it identifies the operating conditions a reader needs in order to understand why different participants may describe the same development differently.
Independent replication and broader task testing are needed before calling the ranking durable. The next documentary tests are third-party evaluations and pricing and actual grok 4.7 timing and meta deployment plans. Until those records appear, the responsible account is bounded: it reports the confirmed action, preserves attribution for contested claims, and does not fill missing evidence with assumptions. That boundary is especially important while the story remains active and decisions are still being made.