xAI Releases Grok Build 0.1 at $1/$2 Per Million Tokens

xAI shipped Grok Build 0.1 on May 21, pricing it at $1.00 per million input tokens and $2.00 per million output tokens. That makes it cheaper than xAI's own Grok 4.3, which launched earlier this month at $1.25 in and $2.50 out, and a fraction of the cost of OpenAI's GPT-5.5 Pro at $5.00/$30.00.

The "Build" designation is the signal worth reading carefully. This is not a flagship competing for benchmark headlines. xAI appears to be creating a separate pricing tier aimed at developers and agentic pipelines, where token consumption compounds fast and cost predictability matters as much as raw capability. Whether the model delivers the latency and reliability those workloads require is still unconfirmed; no independent benchmark results have been published at this writing.

Pricing in context

The positioning makes more sense against the broader release calendar tracked by Price Per Token. Google's Gemini 3.5 Flash debuted the same week at $1.50 in and $9.00 out with a 1M context window. Mistral's Devstral 2 123B, available on Amazon Bedrock since mid-May, undercuts everyone at $0.40/$2.00 for teams willing to work within an open-weight model. Grok Build 0.1 sits between them: more expensive than Mistral's coding-focused option, cheaper than Google's multimodal tier, and deliberately below xAI's own prior release.

That last point is unusual. Labs typically let older flagship models age at higher price points while newer releases come in at a premium. Shipping a cheaper model within weeks of Grok 4.3 suggests xAI is either segmenting its market deliberately or responding to developer feedback that 4.3's rates were too steep for high-volume inference.

The developer market

This release lands at an inflection point in the artificial intelligence market. Analytics Insight reports that Anthropic overtook OpenAI in paid business adoption for the first time in April 2026, capturing 34.4% to OpenAI's 32.3% according to Ramp's AI Index, which tracks spending across more than 50,000 companies. Gartner projects global AI model spending to approach $33 billion by year-end.

LLM Stats illustrates the pace of competition: five significant launches in May alone, spanning Google, OpenAI, Meta, and xAI. In a market this crowded, pricing has become a genuine competitive axis for teams running repeated inference at scale. A $1.00 versus $1.50 input rate is a 33% difference that compounds into thousands of dollars monthly for high-frequency API users, making a rigorous artificial intelligence review of unit economics a routine procurement step now.

What we do not know

xAI has not released Grok Build 0.1's context window size, latency figures, or structured evaluation results comparing it to Grok 4.3. The model name implies developer orientation, but practitioners should run domain-specific evals before committing workloads to a cost model that only pays off if quality holds up. Pricing is the easy variable to verify; capability is not.

This is worth contrasting with how Anthropic handled its Mythos model. As PBS NewsHour reported, Anthropic gave access to over 40 companies specifically to stress-test the system before any wider release, prioritizing vulnerability identification over speed. xAI's approach with Build 0.1 appears to invert that sequence: ship first, gather real-world signal, iterate. Neither posture is inherently wrong, but the risk profile differs sharply depending on the deployment context and how much your application tolerates unexpected model behavior.

Neither approach dominates unconditionally. The choice reflects organizational philosophy as much as product strategy.

The market reaction

Adoption for Grok Build 0.1 will hinge on factors pricing alone cannot settle: API reliability, rate limits, uptime SLAs, and the consistency that production agentic systems demand. xAI has demonstrated it can ship at pace. The more pointed question is whether Build 0.1 represents a sustained commitment to the developer tier or a tactical pricing move to pressure competitors in a week when Gemini 3.5 Flash was getting attention.

With model spending accelerating and the market fragmenting into capability and cost tiers, a lab that credibly serves both enterprise and high-volume developer needs holds a structural advantage. xAI undercutting itself may be the most honest signal yet about where it thinks the next wave of AI adoption is actually coming from.

---

FAQ

Q: What is Grok Build 0.1?
A: It is a new model from xAI released on May 21, 2026, priced at $1.00 per million input tokens and $2.00 per million output tokens, available through xAI's API.

Q: How does Grok Build 0.1 pricing compare to Grok 4.3?
A: Grok Build 0.1 is approximately 20% cheaper on both input and output than Grok 4.3, which launched earlier in May at $1.25/$2.50 per million tokens.

Q: How does Grok Build 0.1 compare to GPT-5.5 Pro pricing?
A: GPT-5.5 Pro is priced at $5.00 in and $30.00 out per million tokens, making Grok Build 0.1 five times cheaper on input and fifteen times cheaper on output.

Q: Is Grok Build 0.1 better than Grok 4.3 on benchmarks?
A: No independent benchmark results have been published for Grok Build 0.1 as of this writing. Its name suggests a developer-focused use case, but capability comparisons remain unconfirmed.