Meta, OpenAI and SpaceXAI Launch New AI Models With Lower Prices

Meta opened its first paid model API on July 9 and priced Muse Spark 1.1 at $1.25 per million input tokens and $4.25 per million output tokens. OpenAI made the GPT-5. SpaceXAI shipped Grok 4.5 on July 8.

DeepSeek made a 75 percent cut to V4-Pro pricing permanent in late May. Forbes reported that three near-frontier launches occurred inside eight days. OpenAI keeps a 90 percent discount on cache reads for GPT-5.6 and bills cache writes at 1.25 times the uncached input rate.

Meta reads cached input at $0.15 per million tokens. Anthropic charges separately for cache writes at 1.25 times input for the five-minute window and twice input for the one-hour window. Muse Spark 1.1 bills its internal thinking tokens at the output rate.

GPT-5.6 offers an ultra setting that coordinates four agents in parallel. OpenAI audited SWE-Bench Pro and estimated that roughly 30 percent of its tasks were broken. The company said it no longer recommends the benchmark as a leading coding eval.

Goldman Sachs forecasts that token consumption will multiply 24 times between 2026 and 2030, reaching 120 quadrillion tokens per month. Studies of agentic coding report per-task usage orders of magnitude above simple chat and wide variance between repeated runs of an identical task.

Agentic systems turn a single user request into repeated rounds of planning, retrieval, tool calls, validation and retries.

Forbes reported that input tokens drive the total in agentic coding workloads because the system resends repository context, tool results and conversation history on every turn. Total organizational spending multiplies task volume by attempts per task by tokens per attempt by effective token price, plus tool and infrastructure costs. Only the price term is falling.