SpaceXAI has unveiled Grok 4.5, a new model designed for programming, agent tasks, and intelligent work. The company claims it is their most powerful system to date.
Announcing Grok 4.5, our first model trained specifically for coding and agents. It was trained with Cursor and offers frontier intelligence at leading speeds and cost efficiency.https://t.co/i8HpU7w64k pic.twitter.com/oBjGtTsoNc
— SpaceXAI (@SpaceXAI) July 8, 2026
The model is available via API, in Grok Build, and Cursor across all pricing tiers. However, Grok 4.5 is not yet available in the European Union, with a launch expected in mid-July.
The main focus of this release is cost and speed. Grok 4.5 is priced at $2 per million input tokens and $6 per million output tokens. In comparison, Anthropic's Claude Opus 4.8 costs $5 and $25 respectively, Claude Fable 5 costs $10 and $50, while OpenAI's GPT-5.6 Sol is priced at $5 and $30.
GPT-5.6 is currently in limited preview mode: OpenAI has opened the model family to a small group of trusted partners and plans to make Sol, Terra, and Luna publicly available in the coming weeks. Within the same lineup, Luna has a lower price of $1 per million input tokens and $6 per million output tokens.
According to Elon Musk, Grok 4.5 is an "Opus-class model, but faster, cheaper, and more token-efficient." He later clarified that, based on SpaceXAI's internal assessment, the system is "roughly comparable to Opus 4.7, but much faster."
Based on strong positive feedback from customers in our beta test program, @SpaceXAI will make Grok 4.5 available to the public tomorrow.
It is an Opus-class model, but faster, more token-efficient and lower cost.
— Elon Musk (@elonmusk) July 8, 2026
Test Results
According to published results from SpaceXAI, Grok 4.5 has shown mixed performance: it did not lead in all tests but also did not fall significantly behind its competitors.
In the DeepSWE 1.0 test, Grok 4.5 scored 62%. It trailed behind Fable at 66.1% and GPT-5.5 at 64.31%, but outperformed Claude Opus 4.8 at 55.75% and Opus 4.7 at 40.12%.
Source: SpaceXAI.In the DeepSWE 1.1 test, the score was lower: Grok 4.5 achieved 53% compared to 59% for Opus 4.8, 67% for GPT-5.5, and 70% for Fable. On the SWE Bench Pro, the model scored 64.7%, which is higher than GPT-5.5's 58.6% and nearly on par with Opus 4.7's 64.3%, but lower than Opus 4.8's 69.2% and Fable's 80.4%.
In the Terminal Bench 2.1, Grok 4.5 scored 83.3%, almost matching GPT-5.5's 83.4%, but trailing Fable's 84.3%. In the SWE Marathon, it took first place with 29% against 26% for Opus 4.8 and 24% for Fable. The company also reported Grok 4.5's leadership in Harvey's Legal Agent Benchmark.
SpaceXAI's Argument
The stronger argument for this release is not based on peak test scores but on economics. According to SpaceXAI, in SWE Bench Pro tasks, Grok 4.5 used an average of 15,954 output tokens per task. In contrast, Claude Opus 4.8 had a figure of 67,020 tokens—4.2 times more.
Source: SpaceXAI.For teams running AI on a large volume of tasks, this impacts the overall cost of operations. With a lower price for output tokens, Grok 4.5 may be more cost-effective in scenarios with many iterations: bug fixes, generating edits, code reviews, and agent cycles.
SpaceXAI also reported a generation speed of around 80 tokens per second and a context window of 500,000 tokens. The company positions the model as a compromise between quality, speed, and cost.
Connection with Cursor
In the release, SpaceXAI stated that Grok 4.5 was trained alongside the AI service Cursor. The company did not disclose the full dataset but indicated that the model was developed for programming tasks, agent work with tools, science, engineering, and mathematics.
According to Decrypt, the training utilized data from Cursor developer sessions, including debugging traces and actual code edits, rather than just static repositories.
In June, SpaceX entered into an agreement to acquire the AI service valued at $60 billion. The deal will be conducted in SpaceX Class A shares, with closure expected in Q3 2026 pending regulatory approvals.
Grok 4.5 is one of the first major models following Musk's integration of the AI division with SpaceX. The company utilized tens of thousands of Nvidia GB300 GPUs for its training.
It is worth noting that in February, media reported on plans by the then-separate SpaceX and xAI to develop software for the Pentagon's autonomous weapons.