DEV Community: Steriani Karamanlis

Platforms Cross Below Neoclouds in 2026 First as Cached Pricing Diverges 59 Points Between Frontier and Platform Channels YTD

Steriani Karamanlis — Tue, 02 Jun 2026 14:26:51 +0000

Last week we wrote that twenty weeks of clean data had surfaced a two-track market, with text compressing 9.4 percent year to date while reasoning, mid tier, and budget output ran double-digit gains. Week 21 layers a second structural dimension on top of that map. The channel stack reshuffles, and for the first time in 2026 third-party inference platforms now price close to or even below GPU-native neoclouds for the average text model.

Platform discount widened from 48.0 percent in Week 20 to 50.1 percent in Week 21. Neocloud discount tightened from 50.0 percent to 48.6 percent. The two channels have crossed, reversing a discount ranking that held for the entire first quarter of the year. Cloud marketplaces still sit furthest from direct developer pricing, but the relationship between the two specialty channels has changed shape.

Size spread compressed from 7.2x to 6.2x. The premium that larger models commanded over smaller alternatives in Week 20 receded toward the 5.8x reading of Week 19. The structural KPIs continue to do the visible work while the price benchmarks themselves hold steady.

The week itself was quiet, and the channel reshuffle did the talking

Week 21 was again quiet in the matched-set. Most of the fifteen AIPI benchmark indexes posted moves under 40 basis points on every pricing direction. AIPI PLT GLB cached input recorded the largest single move on the week at minus 0.39 percent, reflecting continued aggressive cached repricing across third-party platforms. AIPI CLD GLB output was down 0.14 percent and input down 0.09 percent. AIPI TXT GLB output was down 0.07 percent and input down 0.05 percent. AIPI FTR GLB cached input was down 0.08 percent, and the rest of the frontier flagship benchmark held within two basis points across input and output. The chained matched-model methodology absorbed 5,296 priced SKUs from 51 vendors with the year-to-date picture intact and the weekly variance contained.

The story of Week 21 is not in the indexes. It is in the channel discounts, the cached pricing divergence, and the size spread retracement that all moved together while individual SKU prices stayed close to their Week 20 levels.

Why the channel stack crossed

The platform discount widened because third-party inference platforms continued to capture cheaper open weight models and to expand cached pricing tiers. Across the 696 SKUs in the platform channel, average input pricing now sits at $0.000680 per 1,000 tokens. Direct model developer pricing sits at $0.004702 per 1,000 tokens across 936 SKUs. The numerical advantage was already in platform’s favor on input pricing earlier in the year. What changed in Week 21 is that the structural KPI caught up to the underlying SKU pricing reality and crossed the neocloud line.

The neocloud discount tightened from a different direction. Neocloud pricing held remarkably stable, with AIPI NCL GLB at $0.000301 per 1,000 tokens on input and minimal weekly movement across 122 SKUs. What shifted is the model developer baseline against which the discount is measured. As AIPI DEV GLB drifted modestly to minus 0.7 percent YTD on input, the gap between neocloud and direct developer narrowed. Neocloud did not get more expensive. Direct developer pricing got cheaper while neocloud held.

The cached pricing split

AIPI FTR GLB cached input sits 39.2 percent above January. AIPI PLT GLB cached input sits 19.5 percent below. The 59-point spread between frontier and platform channels is the widest YTD divergence in any cached pricing direction since tracking began in December.

Frontier flagship models entered 2026 with cached input pricing structured as a 70 to 90 percent discount off non-cached input. Through Q1 and into Q2, several vendors restructured cached tiers, in some cases tightening the discount and in others bundling cached input with new prompt processing minimums. The result is that the chained matched-model frontier cached benchmark has climbed in steps over twenty-one weeks, not in a single move.

AIPI PLT GLB cached input moved in the opposite direction. Inference platforms expanded cached access aggressively through Q2, with caching availability across all text models crossing 23.4 percent of the population in Week 21, up from 22.8 percent in Week 19 and 16.9 percent at the start of Q2. The pricing tiers offered through platforms tend to be discounted below direct developer cached pricing as a category, and the cumulative effect is a 19.5 percent YTD decline in the platform cached benchmark.

The 59-point divergence is not a single-week event but rather the cumulative footprint of three months of structural moves that pulled frontier cached pricing up and platform cached pricing down at the same time.

The two tracks hold

The two-track YTD picture from last week holds intact. AIPI TXT GLB sits 9.4 percent below January on input. AIPI MID GLB carries 13.0 percent on input and 14.5 percent on output. AIPI RSN GLB runs 11.4 percent on input and 10.1 percent on output. AIPI BDG GLB output is up 12.4 percent year to date. AIPI FTR GLB stayed close to flat at minus 0.2 percent on input and minus 0.1 percent on output, completing the picture from last week that frontier flagship is the only tier holding its January reference within a basis point band.

Open source advantage widened from 68.8 percent in Week 20 to 69.7 percent in Week 21. Context cost curve ticked up from 3.4x to 3.6x as long context model pricing repositioned at the upper end. Reasoning premium compressed from 1.8x to 1.7x, continuing the gradual tightening from 2.2x in Week 17. Caching availability crossed 23.4 percent of text models.

Coverage and structure

Week 21 coverage settled at 5,296 SKUs across 51 vendors and 3,249 models, modestly below Week 20 as nine vendors completed planned delistings of legacy models flagged through their April change-detection logs. The delistings touched the budget and mid tier segments most heavily, but the matched-model methodology absorbed them without disturbing the chained signal.

The channel population physics clarifies how the crossover took shape. Platform covers 696 SKUs, neocloud covers 122, and direct model developers cover 936. Platform is roughly six times the size of neocloud by SKU count, and that breadth is what enabled platform pricing to find new equilibria more quickly than the smaller neocloud pool. Neocloud, with a thinner SKU base, exhibits less monthly variance and tends to hold pricing for longer windows. The compression in the platform channel had room to run, and it ran into the neocloud line in Week 21.

The cached pricing split traces back to a similar population dynamic. Of the 800 text models tracked in Week 21, 187 offer cached input pricing, or 23.4 percent of the universe. Those 187 SKUs split unevenly across channels, with platforms holding the deepest cached pricing pool relative to their direct developer counterparts.

Forward look

The first Q2 quarterly read becomes possible next week. If the channel crossover holds through Week 22, it becomes the leading framework for that analysis. If it reverts, Q2 stays anchored on the two-track tier narrative the index has carried since April.

DeepSeek V4 propagation through third-party platforms continued through Week 21. If the series lands at lower channel pricing than the V3 models it is replacing, AIPI PLT GLB cached could extend its 19.5 percent YTD decline further and widen the channel cached spread beyond 60 points.

The Grok 4.2 alignment on Azure Foundry that landed in Week 21 brought frontier flagship channel pricing closer to a unified rate across providers. Week 22 will test whether other Azure-hosted frontier model series follow with similar price alignments, particularly in cached input where the channel spread remains wide.

Methodology and Resources

The Inference Price Benchmark publishes every Monday at 9am Eastern Time.

Full index methodology is available at a7om.com/methodology.

Live pricing for AI agents and dev workflows: a7om.com/mcp

Full inference market intelligence for analysts and FinOps: a7om.com/terminal

Structured data feeds for enterprise licensing: a7om.com/feed

Get AIPI Weekly direct to your inbox every Monday: https://a7omintelligence.substack.com/subscribe

We Publish a Free Weekly AI Inference Pricing Index. Here Is How To Get It.

Steriani Karamanlis — Fri, 15 May 2026 09:30:59 +0000

Every Monday, we publish the ATOM Inference Price Benchmark, a free weekly index tracking per-token pricing across 51 AI inference vendors, 5,000+ SKUs, 3,000+ models and 9 countries.
It is the only chained matched-model inference price index published publicly. The methodology is deterministic and zero variance. No opinions. Just data.
What you get every Monday:
15 AIPI indexes across modality, channel, tier and geography. 9 market KPIs including Open Source Advantage, Reasoning Premium, Caching Discount and Channel Spread. Forward calls on pricing movements before they happen. Full coverage across text, image, audio, video, embedding and reasoning modalities.
This week we called a 17.47% drop in platform cached input pricing two weeks before it happened. The forward call resolved in exactly the column we flagged.
If you build with AI, buy inference at scale, or track the economics of the AI infrastructure market this is the one data source worth following.
Subscribe free here: linkedin.com/build-relation/newsletter-follow?entityUrn=7455608005699534848
Full dataset and live pricing at a7om.com

First Confirmed Directional Move on the AI Inference Frontier Index in 2026

Steriani Karamanlis — Tue, 12 May 2026 15:03:26 +0000

By Stamos Kanellakis, Founder of ATOM

For the past 17 weeks I've been tracking per-token pricing across 51 AI inference vendors and 5,000+ SKUs. This week the index posted something we haven't seen all year: a confirmed directional move on the frontier.

The numbers are small. The shape is unusually clean.

What the data shows

The frontier index (AIPI FTR GLB, covering peak-capability flagship models like Claude Opus 4.7, GPT-5.5, Gemini 3.1 Pro) declined for the third consecutive week:
Input: -0.23%
Cached input: -2.06%
Output: -0.35%

The output figure is nearly identical to last week, which is part of what makes the trend look real.

The global text benchmark (AIPI TXT GLB, which covers the full text-generation market across all tiers) moved with it. For the first time in 2026:
Input: -0.35%
Cached input: -1.01%
Output: -0.23%

The pattern is no longer confined to flagship models. It's now visible across the wider text market.

Why three weeks matters

Single-week moves on the frontier index are common in size but usually random in direction. Vendors reprice individual SKUs without coordinating with each other, which produces noise inside a tight range.

Two weeks down starts to feel like something. A third week makes it hard to explain as noise, and easier to explain as several frontier vendors pulling in the same direction.

Week 18 adds two more reasons to take the signal seriously:

The same pattern now appears on the global text benchmark, which covers many more models than the frontier subset. Getting that index to shift requires either a large share of vendors moving together, or a handful of heavy vendors pulling the rest along.
Input, cached input, and output all softened at once on the same index in the same week. Vendors rarely move all three columns together when they are only running targeted promotions.

The forward call resolved

Two weeks ago I flagged two scheduled events that should show up in this week's run:

DeepSeek's 75% promotional discount on V4-Pro
Alibaba Cloud Bailian's cut to cache pricing for DeepSeek V4-Pro (RMB 1 per million tokens)

Both changes touched cached pricing rather than headline rates. So the impact should appear in the cached input column of the platform channel.

That is exactly where it landed:
AIPI PLT GLB (Platform channel)
Cached input: -17.47% ← largest single move on any AIPI series this week
Input: -0.71%
Output: -1.08%

A forward call resolving in the exact column we flagged is what separates a price tracker from a benchmark.

Coverage

Models: 3,079
SKUs: 5,022
Vendors: 51
Countries: 9
Modalities: 6 (with 35 subtypes)

The 190 SKU increase from last week came mostly from continued catalog growth in audio, voice, and image generation.

Reasoning models corroborate

AIPI RSN GLB input declined 0.50%, and the reasoning premium KPI compressed from 2.2x to 1.7x.

Part of that compression comes from new reasoning entrants joining at lower price points rather than incumbents cutting rates, so the change is not a clean price signal on its own. What matters more is that reasoning is softening at the same time as the flagship segment. Reasoning is a frontier capability, and the two indexes have historically tracked together.

What's becoming calmer vs what's moving

Volatility across the wider market continues to drop:
Input volatility YTD: 0.61% (-0.34pp from Week 17)
Cached input volatility YTD: 0.30% (-0.44pp from Week 17)
Output volatility YTD: 0.45% (-0.45pp from Week 17)

The broader market is becoming calmer at the same time the frontier subset is starting to move. That is an unusual combination and worth watching.

What to watch for Week 19

Three items are likely to shape Week 19:

Whether the pattern on AIPI FTR GLB extends to a fourth week. If it does, it becomes the longest sustained directional run on the index since we started publishing.
xAI's scheduled retirement of grok-imagine-image-pro on May 15. This falls inside the Week 19 indexing window. Two more retirements are already on the calendar: Moonshot's original Kimi K2 series on May 25, and the Writer Palmyra-x-003 family on July 13.
Audio segment. AIPI AUD GLB recorded zero movement on all three pricing directions this week after the 5.77% input jump in Week 17. The segment looks like it has settled at a new baseline of 223 SKUs. The question for Week 19 is whether audio stays calm or whether new entrants pull it back into volatility.

Methodology note

ATOM indexes use a chained matched-model methodology. Only SKUs present in both the current week and the prior week contribute to the weekly percent change, which removes the composition bias that affects simple average pricing. A maximum weekly cap of ±50% is applied at the SKU level to prevent outlier movements from distorting the aggregate.

The full Week 18 edition of The Inference Price Benchmark, including additional breakdowns by channel and modality, is published every Monday at 9am ET. Read the full edition here.

Track per-token pricing across 51 AI inference vendors at a7om.com.

OpenAI just raised $122B. Frontier inference pricing hasn't moved in 9 weeks

Steriani Karamanlis — Fri, 03 Apr 2026 15:12:54 +0000

OpenAI just closed the largest venture round in history at $852B valuation. Record capital, record confidence in AI's future.
But here's what's interesting from a market pricing perspective. Frontier model pricing has been completely flat for 9 consecutive weeks. The benchmark sits at $0.005714 per 1K input tokens across top tier flagship models globally.
At the same time the spread between frontier and budget models is 7.1x. That's a significant gap that's been holding steady.
So the question the market is now asking is which way does this go from here. Does record capital give frontier labs room to hold pricing while budget models keep improving? Or does the competitive pressure eventually compress the premium?
For teams building on top of inference at scale this dynamic matters a lot. The model selection decision isn't just a capability question anymore, it's a cost strategy question.
Curious what others think. Are you seeing this pressure in your own stack decisions?
We publish weekly inference pricing intelligence at a7om.com if you want the underlying data.

Query Live AI Inference Pricing with the ATOM MCP Server

Steriani Karamanlis — Thu, 26 Mar 2026 20:48:24 +0000

If you've ever tried to compare LLM pricing across vendors you know how painful it is. One charges per token, another per character, another per request. Cached input discounts exist but good luck finding them. Context window pricing is buried. And by the time you've normalized everything into a spreadsheet something changed on a pricing page and your numbers are stale.

This is the problem ATOM was built to solve. It tracks 2,583 SKUs across 47 vendors, normalizes everything to a common unit, and exposes it all through an MCP server your agents can query directly.

Here's how to set it up and what you can actually do with it.

What MCP gives you here

Model Context Protocol lets AI agents connect to external data sources through a standardized interface. Claude, Cursor, Windsurf and others support it natively.

Instead of pasting a pricing table into your prompt and hoping it's current, you give your agent a live connection to the source. It queries, reasons, and acts on real numbers.

Setting up the ATOM MCP server

ATOM's server is published on npm, Smithery, and the official MCP registry.

Claude Desktop

Add this to your claude_desktop_config.json and restart:

{
  "mcpServers": {
    "atom-pricing": {
      "command": "npx",
      "args": ["-y", "atom-mcp-server"]
    }
  }
}

Cursor or Windsurf

Add the server endpoint in your MCP settings:

https://atom-mcp-server-production.up.railway.app/mcp

Any other MCP client

The server supports both HTTP SSE and stdio transport. Run it locally via npx or point at the Railway endpoint above.

The tools

The free tier includes 4 tools that give you macro market intelligence with no login required. MCP PRO ($49/mo) unlocks the remaining 4, which give you model-level and vendor-level detail.

Free tier

Tool	What it does
`list_vendors`	All 47 tracked vendors with type and region
`get_kpis`	6 live market KPIs updated weekly
`get_index_benchmarks`	14 AIPI price indexes by modality and tier
`get_market_stats`	Aggregate supply and cost structure data

MCP PRO

Tool	What it does
`search_models`	Filter by context size, tool support, modality, price
`get_model_detail`	Full spec and pricing for a specific model
`compare_prices`	Cross-vendor comparison for a model family
`get_vendor_catalog`	Full SKU list for a specific vendor

What it looks like in practice

Check what the market looks like right now (free)

get_kpis

This week's numbers:

Output tokens cost 3.84x more than input tokens on average
Cached input saves 69.7% vs standard input pricing
Open source models run 80% cheaper than closed source equivalents
Only 20.3% of SKUs in the index offer cached pricing at all
The price gap between small and large models in the same family is 4.8x

These are median figures across all tracked SKUs, recalculated every Monday.

Find the cheapest model with 100K+ context and tool calling (PRO)

search_models
context_window_min: 100000
tool_calling: true
sort_by: input_price_asc

Returns model-level results with normalized per-token pricing across vendors. The spread between cheapest and most expensive for functionally similar models is typically over 30x. That difference compounds fast at any real usage volume.

Compare vendors for a specific model family (PRO)

compare_prices
model_family: "Llama 3.3 70B"

Returns every vendor offering that model with normalized pricing so you can make a direct comparison without doing any unit conversion yourself.

Why this is useful for agent architecture

If you're building anything that makes a lot of LLM calls, model routing based on cost and capability is a real decision you're making, consciously or not. The cheapest model that can handle a task should handle it.

With ATOM connected your agent can check current prices before picking a model, catch when a vendor changes pricing, estimate the cost of a planned workload before running it, and compare vendors for a specific capability requirement. That reasoning used to mean a spreadsheet someone had to maintain. Now it's a tool call.

A note on the data

ATOM uses a chained matched-model methodology, the same logic you'd apply to a commodity price index. Every SKU is normalized to a common unit, timestamped, and verified. The point of the methodology is to eliminate composition bias so week-over-week comparisons are actually meaningful and not just reflecting which vendors got added or dropped.

Full methodology at a7om.com/methodology.

Try it

Run npx atom-mcp-server or search "ATOM" on Smithery. Free tier covers 4 tools with no login. MCP PRO is at a7om.com/mcp.

The inference market now has a benchmark. Might as well use it.

"H, I just joined DEV. I've spent the past year building ATOM, a live pricing benchmark for AI inference. Tracks 2,500+ SKUs across 47 vendors. First article dropping this week on querying it via MCP. Follow along if that's relevant to what you build."

Steriani Karamanlis — Thu, 26 Mar 2026 20:13:32 +0000