The Aggregate — LLM benchmark aggregate

Measured conclusion: Deep-tested. L2 outcome: tools_ok.

LLM rankings from public benchmarks, model comparisons and source-linked results. Updated daily.

Measured statistics

Add to your agent

Copy this into your MCP client config:

{
  "mcpServers": {
    "The Aggregate — LLM benchmark aggregate": {
      "url": "https://theaggregate.ai/mcp",
      "transport": "streamable-http"
    }
  }
}

Measured tools (8)

• get_leaderboard read-only

  Top of the cross-benchmark aggregate ranking: every model placed on one rating scale by an IRT model fit over public benchmark leaderboards (call about_the_aggregate for the current coverage counts). One row per model by default, fused across reasoning-effort settings. Supports paging via limit/offset.

  Parameters:

  • limit — number — Rows to return (1-100, default 25).
  • offset — number — Rows to skip from the top (default 0).
  • include_variants — boolean — Rank each reasoning-effort variant separately (e.g. "Claude Opus 4.6 (High)") instead of one fused row per model. Default false.

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• search_models read-only

  Find ranked models by (partial) name or provider. Returns rank, Aggregate Rating and the model page URL. One row per model by default, fused across reasoning-effort settings.

  Parameters:

  • query (required) — string — Model or provider name fragment, e.g. "opus" or "deepseek".
  • limit — number — Max results (1-25, default 10).
  • include_variants — boolean — Return each reasoning-effort variant separately (e.g. "Claude Opus 4.6 (High)") instead of one fused row per model. Default false.

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• get_model read-only

  One model in depth: aggregate rank, Aggregate Rating with standard error, provider, what it is, cost per task where known, and its most notable benchmark results (with percentiles).

  Parameters:

  • model (required) — string — Model name or slug, e.g. "Claude Opus 4.5" or "gpt-5-5".

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• compare_models read-only

  Head-to-head between 2-4 models: aggregate ranks, rating gap with a significance note based on the standard errors, and notable benchmarks they share.

  Parameters:

  • models (required) — array — Two to four model names or slugs.

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• search_benchmarks read-only

  Find benchmarks in the aggregate by (partial) name. Returns model coverage, difficulty on the rating scale, and the benchmark page URL.

  Parameters:

  • query (required) — string — Benchmark name fragment, e.g. "swe-bench" or "arena".
  • limit — number — Max results (1-25, default 10).

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• get_benchmark read-only

  One benchmark in depth: what it measures, the original source leaderboard URL, IRT stats (difficulty, noise, model coverage), skill weights, and the current top models on it.

  Parameters:

  • benchmark (required) — string — Benchmark name or slug, e.g. "Aider polyglot".
  • top — number — How many top models to list (1-50, default 10).

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• get_prediction_duel read-only

  Guesswork — the public prediction duel: every day frontier LLMs and The Aggregate's own IRT model predict newly scraped benchmark scores before seeing them, and the errors are scored. Returns the current monthly standings, wins and losses included.

  Hints: read-only yes · destructive no · idempotent yes · open-world no

• about_the_aggregate read-only

  What this data is: how the IRT fusion works, current coverage counts, update cadence, and how to cite it.

  Hints: read-only yes · destructive no · idempotent yes · open-world no

Publisher website

Measured by MCPYP probing. Untested fields are omitted, never estimated. Listing is always free; MCPYP never proxies calls or stores user tokens.