The spot market for AI Inference
Joined September 2025
- Tweets244
- Following87
- Followers1.4K
- Likes280
The Grid retweeted
We need a token trading market. Trade token contracts for each model. Who is doing that?
The Grid retweeted
Live now on @The_GridAI
GPT 6.1 Sol
Simply hit gpt-sol-latest
As always, -latest instruments fall back to previous versions for reliability
The Grid retweeted
Carl DiClementi (@CahlDee) joins the stream tonight to introduce @The_GridAI, which lets you buy AI model usage at a price set by competing providers instead of whatever one company charges.
youtube.com/watch?v=K2Y0z2LO…
The Grid retweeted
A lucky buyer today got 11 million tokens of inference at $0.0001 per million. Kinda wishing I had a limit order there to catch it.
Text Prime is currently serving MiniMax-M3.
That's >1000x cheaper than you'll find anywhere else.
Yes, with zero data retention included. Your data is your own.
Name another place where you can buy the dip to save on tokens.
The Grid retweeted
You don't need to change your existing OpenAI or Anthropic setup. Just update the base URL and replace the model name with text-prime. That's it.
The Grid retweeted
Uber reportedly torched $3.4B of its AI budget in just four months.
The root cause: an absolute lack of a routing layer between the request and the model.
We are blindly throwing expensive frontier models at EVERY SINGLE TASK.
This is why I am excited about @The_GridAI.
It completely flips the script by turning AI inference into a live spot market.
Instead of hardcoding a single, expensive provider, you select a quality tier eg:
- text-standard
- text-prime
- text-max
The Grid then dynamically routes your request to the cheapest qualifying model 🔥
→ Drop-in migration via an OpenAI-compatible endpoint
→ Zero vendor lock-in for your applications
→ Automated, cost-saving routing based on task difficulty
→ Complete visibility into every token spent
Just for fun, I built a small @Streamlit demo to showcase its capabilities.
The app lays out the entire process:
✦ Header & Request:
↳ a clean UI to plug in an API key and choose your instrument (text-standard, text-prime, or text-max).
✦ Routing & Metadata:
↳ run a prompt and instantly inspect the backend
You can see the latency, prompt/completion tokens, and the exact model returned.
✦ Migration & Traceability:
↳ a live look at the code swap and how the market actually routed the request.
On the top left, you can enter your API key and try it out for yourself!
@The_GridAI just went live, and they are giving new accounts an incredible 200 million free tokens.
Demo app + open-source repo in the 🧵↓
🤝 Paid partnership
The Grid retweeted
You don’t need to keep moving inference around manually - suppliers compete while the routing layer handles the switching...
The Grid retweeted
Genius idea for AI inference!
A marketplace that routes requests to the cheapest qualifying model at any given point.
This can get you up to 87% cheaper inference!
Today, if you need a model, you pay the vendor's fixed rate card, but that's about to change with this:
The Grid retweeted
🧵 3. The 3 text tiers and the 3 pricing
- Standard is for high-volume work where cost matters most, like classification, batch summarization, tagging, and simple extraction.
- Prime is the everyday production tier. This is where I’d put agents, RAG, drafting, support workflows, and quality-sensitive pipelines.
- Max is for the harder stuff, like long-context work, high-stakes reasoning, and tasks where a wrong answer can create real downstream cost.
The important part is that “cheapest” does not mean “random cheap model.”
Each tier has a quality threshold. The Grid checks models against benchmark floors anchored to Artificial Analysis. If a supplier falls below the required quality level for a tier, it gets removed from the eligible set. So the market competes on price, but only inside the quality bar you picked.
The Grid retweeted
Tokens are becoming units of machine labor. Premium tokens are just higher grade labor like harder tasks or better outputs. If they're commoditized then we need a commodity style market @The_GridAI
The Grid retweeted
If AI sells work itself then the price of work won't anchor to human wages. It starts anchoring to marginal inference cost. Those costs are almost definitely going to continue falling and will lead to demand explosion into work that was never economical before
There's a new world of labor markets that are gonna need price discovery and @The_GridAI
The Grid retweeted
Token consumption go up
Need real time price discovery and physically metered delivery of llm tokens
Need commodity style markets with standardization
Need @The_GridAI
I read Goldman Sachs’ AI report, and I was genuinely impressed.
The core insight is as follows:
Agentic AI could turn AI from a capex-heavy cost burden into a business where usage growth drives margin expansion. As token costs fall, more complex agents become economically viable. These agents then consume far more tokens through longer context windows, repeated reasoning loops, validation, tool use, and always-on background monitoring.
This increase in token usage improves infrastructure utilization, strengthens unit economics, and gives hyperscalers and model providers more room to reinvest in model quality, distribution, and capacity.
In other words, the bull case for AI capex is not simply that usage will grow. It is that this usage growth can increasingly flow through at attractive incremental margins. Goldman Sachs argues that this margin inflection is beginning to appear from 2026 onward.
The Grid retweeted
At a very different scale, it's the same at @The_GridAI. Everyone does what they do best plus some more. "Agency/Ownership/Competence" could cover most job descriptions. Titles, if used, are basically an API for the outside world.
This quoted post is unavailable.
The Grid retweeted
You wouldn’t use the same grade of steel for a bridge and a bicycle frame.
Same goes for inference.
That’s why we defined three tiers:
▫️Text Max (new) → frontier reasoning for complex tasks
▫️Text Prime → production workloads (quality + speed)
▫️Text Standard → high throughput, lowest cost
Each tier is defined by performance, not model names.
You choose the outcome.
The market competes to deliver it.
The Grid retweeted
fixed api pricing was always a placeholder
real time markets for intelligence coming soon
The Grid retweeted
Once LLM quality is "good enough", most use cases will lean toward speed and cost optimizations.
Google gets this.
Google just priced intelligence at $0.25 per million input tokens. Let that math sink in.
Gemini 3.1 Flash-Lite costs 4x less than Claude 4.5 Haiku on input ($0.25 vs $1.00) and 3.3x less on output ($1.50 vs $5.00). It runs 2.5x faster time-to-first-token than Google’s own 2.5 Flash. And it scores 86.9% on GPQA Diamond, which beats larger Gemini models from previous generations.
This tells you everything about where the AI model war is actually being fought right now. Everyone’s watching the frontier models compete on reasoning benchmarks. The real war is in the efficiency tier, where the actual infrastructure bills get paid.
Here’s why. Enterprise AI is at roughly 10% adoption heading toward 50%. The workloads that drive that adoption curve aren’t complex reasoning tasks. They’re translation, content moderation, intent routing, catalog processing. Millions of calls per day where the difference between $0.25 and $1.00 per million tokens compounds into hundreds of thousands of dollars per month.
Google is doing something specific here. They’re using 3.1 Flash-Lite as a wedge to lock developers into the Vertex AI ecosystem on high-volume workloads, then upselling them to 3.1 Pro for complex reasoning at $2.00 per million input. The cascading architecture play: cheap model handles 90% of requests, expensive model handles 10%. Total cost drops by 80%+ versus running everything through a frontier model.
OpenAI sees the same dynamic. GPT-5 Nano is priced at $0.05/$0.40 per million tokens. That’s 5x cheaper than Flash-Lite on input. The efficiency tier is becoming a loss leader for ecosystem capture.
The company that wins the next 2 years of enterprise AI is the one whose cheap model is good enough to run every log file, every customer chat, every moderation call without exhausting the cloud budget. Google just made their bid.