Archive · 2026-07

July 2026,
frozen.

These figures were written once, at month close on 5 Aug 2026, and cannot be edited. Cite them freely: this page will read the same in a year. This is a restatement, filed as a new row that references the original.

Restated (Dispatch 116). Two page-specific definitions diverged from the engine's and reached this frozen month. (1) Benchmark saturation counted the sync's 0.000 'not measured on this instrument' sentinel as a real result — the engine refuses on it — which published the lcr spread as 75.667 across 126 models when the measured figures are 74.000 across 120. (2) 'Providers tracked' and the quality-per-dollar cheapest listing were priced off all host_prices rows including the OpenRouter aggregate pseudo-host, which is not a provider and carries the cheapest input price for 150 of 306 models; the market-structure section already excluded it. Both now use real endpoints only. Same window, same source rows, corrected figures. The superseded row is preserved, never edited.

306

Models tracked

Active entries in the live catalog.

70

Providers tracked

Distinct real providers with at least one live price. Aggregator listings are excluded.

21

Price moves in July 2026

0 up · 21 down. 1134 new listings are counted separately.

Price moves

What changed in July 2026

Recorded from the append-only price ledger — every observed move, per model and per host, on both the input and output side.

21price movesJuly 2026
  • Decreases21
  • Increases0
  • New listings1134
  • Total moves counts increases plus decreases only. New listings are shown here for context and are never folded into that total.
21

Total moves

Increases plus decreases. New listings are not folded in.

0

Increases

21

Decreases

319

New models

First seen in July 2026.

Top 5 increases

No increases recorded this month.

Top 5 decreases

  • qwen/qwen3-30b-a3b-instruct-2507

    OpenRouter

    -0.0%

    blended

    $0.048 $0.048 in (-0.1%) · $0.193 $0.193 out (-0.0%)

  • qwen/qwen3-30b-a3b-instruct-2507

    OpenRouter

    -0.0%

    blended

    $0.048 $0.048 in (-0.1%) · $0.193 $0.193 out (-0.0%)

  • qwen/qwen3-30b-a3b-instruct-2507

    OpenRouter

    -0.0%

    blended

    $0.048 $0.048 in (-0.1%) · $0.193 $0.193 out (-0.0%)

  • qwen/qwen3-30b-a3b-instruct-2507

    OpenRouter

    -0.0%

    blended

    $0.048 $0.048 in (-0.1%) · $0.193 $0.193 out (-0.0%)

  • qwen/qwen3-30b-a3b-instruct-2507

    OpenRouter

    -0.0%

    blended

    $0.048 $0.048 in (-0.1%) · $0.193 $0.193 out (-0.0%)

Who reprices most

Ranked over everything we hold — we began recording price moves on 31 Jul 2026, so this is a short window, not a 90-day history.

  • 1OpenRouter11moves · 1 model
  • 2SiliconFlow10moves · 1 model

Market structure

The same weights cost wildly different money

Identical model, different real provider. Aggregator listings are excluded, so every gap below is a genuine provider-to-provider spread you could act on today.

140

Models on 2+ providers

1

Median providers per model

27

Most providers on one model

154
152%
76
2–326%
47
4–916%
17
10+6%

Providers per model, across 294 models with at least one real (non-aggregator) endpoint. Most weights are single-sourced; a small tail is served everywhere — and that tail is where the spread lives.

  • GPT-5.6 Luna Pro

    openai/gpt-5.6-luna-pro

    $0.050$1.00

    OpenAIAzure · 2 providers · per MTok in

    +1900%

    spread

  • GPT-5.6 Luna

    openai/gpt-5.6-luna

    $0.050$1.00

    OpenAIAzure · 3 providers · per MTok in

    +1900%

    spread

  • DeepSeek V3.2

    deepseek/deepseek-v3.2

    $0.207$3.00

    BaiduSambaNova · 14 providers · per MTok in

    +1348%

    spread

  • gpt-oss-120b

    openai/gpt-oss-120b

    $0.030$0.350

    CoreWeaveCerebras · 17 providers · per MTok in

    +1067%

    spread

  • Llama 3.1 8B Instruct

    meta-llama/llama-3.1-8b-instruct

    $0.020$0.220

    DeepInfraCoreWeave · 5 providers · per MTok in

    +1000%

    spread

  • Gemma 4 31B

    google/gemma-4-31b-it

    $0.090$0.990

    DeepInfraCerebras · 16 providers · per MTok in

    +1000%

    spread

  • Llama 3.3 70B Instruct

    meta-llama/llama-3.3-70b-instruct

    $0.100$1.04

    DeepInfraTogether · 12 providers · per MTok in

    +940%

    spread

  • MythoMax 13B

    gryphe/mythomax-l2-13b

    $0.060$0.450

    NextBitMancer 2 · 4 providers · per MTok in

    +650%

    spread

Quality per dollar

The cheapest model that still clears the band

For each task class we take the leading published score, subtract that evaluation's measured margin, and pick the cheapest model still above the line. Only suites with a real measured margin appear — a benchmark without one cannot back a claim.

  • gpqa

    GPT-5.6 Luna

    openai/gpt-5.6-luna

    Scores 91.10 on aa:gpqa; bar 88.16 (leader 94.10 − margin ±5.94). 35 models clear it. Cheapest listing at OpenAI.

    cheapest qualifying 91.10bar 88.16leader 94.10band width ±5.94 (measured margin)
    $0.050

    per MTok in

  • hle

    Claude Opus 5

    anthropic/claude-opus-5

    Scores 52.60 on aa:hle; bar 51.75 (leader 53.30 − margin ±1.55). 2 models clear it. Cheapest listing at Claude Platform on AWS.

    cheapest qualifying 52.60bar 51.75leader 53.30band width ±1.55 (measured margin)
    $5.00

    per MTok in

  • lcr

    GPT-5.6 Luna

    openai/gpt-5.6-luna

    Scores 74.00 on aa:lcr; bar 65.88 (leader 75.67 − margin ±9.79). 45 models clear it. Cheapest listing at OpenAI.

    cheapest qualifying 74.00bar 65.88leader 75.67band width ±9.79 (measured margin)
    $0.050

    per MTok in

  • scicode

    Gemini 3.1 Pro Preview

    google/gemini-3.1-pro-preview

    Scores 58.90 on aa:scicode; bar 54.99 (leader 60.20 − margin ±5.21). 8 models clear it. Cheapest listing at Google.

    cheapest qualifying 58.90bar 54.99leader 60.20band width ±5.21 (measured margin)
    $1.00

    per MTok in

  • tau_banking

    GPT-5.6 Luna

    openai/gpt-5.6-luna

    Scores 27.22 on aa:tau_banking; bar 26.10 (leader 33.40 − margin ±7.30). 14 models clear it. Cheapest listing at OpenAI.

    cheapest qualifying 27.22bar 26.10leader 33.40band width ±7.30 (measured margin)
    $0.050

    per MTok in

  • terminalbench_v2_1

    GPT-5.6 Luna

    openai/gpt-5.6-luna

    Scores 80.90 on aa:terminalbench_v2_1; bar 78.75 (leader 89.14 − margin ±10.39). 11 models clear it. Cheapest listing at OpenAI.

    cheapest qualifying 80.90bar 78.75leader 89.14band width ±10.39 (measured margin)
    $0.050

    per MTok in

Benchmark saturation

An evaluation stops being usable when the spread between models collapses into the measurement margin. We require the observed spread to exceed twice the margin; a ratio at or below 1.0 means the instrument can no longer tell models apart.

  • tau_banking · aa:tau_banking

    spread 32.37 vs margin ±7.30 across 69 scored models — discriminating.

    2.22×vs 1.0× floor
  • lcr · aa:lcr

    spread 74.00 vs margin ±9.79 across 120 scored models — discriminating.

    3.78×vs 1.0× floor
  • terminalbench_v2_1 · aa:terminalbench_v2_1

    spread 85.77 vs margin ±10.39 across 69 scored models — discriminating.

    4.13×vs 1.0× floor
  • scicode · aa:scicode

    spread 43.20 vs margin ±5.21 across 135 scored models — discriminating.

    4.15×vs 1.0× floor
  • gpqa · aa:gpqa

    spread 59.00 vs margin ±5.94 across 135 scored models — discriminating.

    4.96×vs 1.0× floor
  • hle · aa:hle

    spread 50.50 vs margin ±1.55 across 135 scored models — discriminating.

    16.26×vs 1.0× floor

Embed

Put the live market on your own page

A self-contained frame that rotates the three sharpest numbers of the month: how many prices moved, the steepest rise and the steepest cut. It refreshes itself, needs no script on your site, and always credits CostMyAI.

<iframe src="https://costmyai.com/embed/intelligence-widget"
  title="AI price market — via CostMyAI"
  width="100%" height="200" loading="lazy"
  style="border:0;max-width:520px"
  referrerpolicy="strict-origin-when-cross-origin"></iframe>
  • RotationMonth-over-month move count, biggest rise, biggest cut.
  • FreshnessServer-cached and refreshed every five minutes.
  • SafetyIsolated iframe, no script in your page, nothing configurable.
  • AttributionThe “via CostMyAI” link is part of the widget on every plan.

Archive

Every closed month, frozen and permanently linkable

At 00:00 UTC on the first of each month we write that month's final figures once and never touch them again. A correction is filed as a new restatement row that points back at the original — the number you cited stays exactly as you cited it.

Method

How a switch gets measured

  • Price sync at freeze

    The prices on this page are the catalog as it stood when this month was frozen. The engine itself keeps re-syncing from public provider feeds; the live page reflects that, this archive deliberately does not.

  • Independent benchmark scores

    Quality comes from published third-party evaluations, per task class. We do not run our own private eval and we are never paid for placement.

  • The equivalence band

    A candidate model only qualifies when its score sits inside the band around your current model for the task class in question. Cheaper-but-worse never clears.

  • Measurement margin

    Every score carries its own uncertainty. We compute the real margin and require the gap to survive it before a switch is offered.

  • Latency ceilings

    Median latency is part of the decision, not an afterthought. Set a ceiling and candidates that breach it are dropped before cost is even compared.

  • Refusals with reasons

    When nothing clears, you get a stated reason — not a weaker suggestion. A quiet downgrade would cost you more than the saving is worth.