Catalog Systems Memory benchmarks & evaluation Claude Plays Pokemon (Anthropic)

Claude Plays Pokemon (Anthropic)

https://www.twitch.tv/claudeplayspokemon

Anthropic's public Twitch livestream of Claude (3.7 Sonnet → Claude 4 Opus) playing Pokemon Red. Agent uses screenshot vision + Game Boy controller actions + a scratchpad memory file the model can edit. Demonstrates long-horizon agentic capability publicly; ran continuously from Feb 2025; viewed by hundreds of thousands cumulatively; key data point for Anthropic's marketing around Claude as a long-horizon agent.

Active · Verified 2026-10-28 · fresh as of 2026-10-28 · 25 cited facts, 0 with a third-party source · Suggest a correction

At a glance

Type
Anthropic agent demo — Claude 3.7/4 Sonnet plays Pokemon Red on Twitch vendor fix
Tier
T2
Section
Memory benchmarks & evaluation
Created
2025-02 (Anthropic livestream launch) vendor fix
Latest release
Twitch livestream ongoing; runs distinct per Claude model version vendor fix
License
not applicable — not OSS fix
GitHub
not applicable — no GitHub repo fix
Pricing
not applicable — research demo / community project vendor fix
Funding
not applicable — not commercial fix
Freshness
2026-10-28 fix
Layer
agent-infra repo fix

Memory taxonomy

storage
kv
retrieval
extraction-pull
persistence
session-only
update
agent-controlled
unit
episode
governance
opaque
conflict
agent

When to use

Optimised for: benchmarking long-horizon agentic capability via game play

Anti-fit: not applicable - eval dataset, not a system

Pros & cons

Pros

Public, observable, real-time evidence of long-horizon LLM-agent capability; high mindshare; live discussion of failure modes (loops, forgetting, hallucinating items).

Cons

Closed agent harness (Anthropic-internal); not reproducible by third parties; benchmark protocol informal — no standardized scoring.

Claims & capabilities

Anthropic-run Twitch stream Feb 2025 onwards; Claude 3.7 Sonnet reached Mt. Moon / 2 badges over ~80hr; later runs with Claude 4 Opus reached deeper progress; one of the most-watched public LLM-agent demos of 2025

Technical surface

API surface
not applicable — eval dataset, not a system
Backend storage
not applicable — eval dataset, not a system
Deployment
not applicable — research demo
Embedding model
not applicable — eval dataset, not a system
Multi-tenancy
not applicable — eval dataset, not a system
MCP
not applicable — benchmark / evaluation harness
A2A
not applicable — benchmark / evaluation harness
OpenTelemetry
not applicable — benchmark / evaluation harness

Similar systems

Other memory benchmarks & evaluation in the catalog, ranked by inbound references.

  • LoCoMo T3

    Conversations spanning up to 35 sessions / 300 turns / ~9K tokens, built from LLM-agent personas grounded on temporal event graphs and verified by human annotators. Tests QA (single-hop, multi-hop, temporal, commonsense, adversarial), event-graph summarisation, multimodal dialogue.

  • GAIA T3

    Benchmark for general AI assistants made of real-world questions that require reasoning, multimodality, web browsing and tool use, with a public leaderboard on Hugging Face. Traction: ~917 likes on the Hugging Face dataset

  • LongBench T3

    Tsinghua THUDM. 21 datasets across 6 categories (single-doc QA, multi-doc QA, summarisation, few-shot ICL, synthetic, code completion) in English + Chinese. ~6.7K words English / 13.4K chars Chinese average. v2 at ACL 2025 with harder, longer instances.

  • LongMemEval T3

    5 core memory abilities — info extraction, multi-session reasoning, temporal reasoning, knowledge updates, abstention — via 500 curated questions in synthetic chat histories. Two scales: LongMemEval-S (115K tokens), LongMemEval-M (up to 1.5M).

  • RULER T3

    NVIDIA. Extends NIAH with multi-needle retrieval, multi-hop entity tracing, aggregation, QA — configurable for arbitrary length / difficulty. Evaluated 17 long-context models on 13 tasks.

  • InfiniteBench (∞Bench) T3

    Tsinghua / OpenBMB. First benchmark with average input >100K tokens. 12 tasks across English + Chinese: long novel QA, code debugging, math in long context, summarisation, retrieval. Realistic documents (novels, code) rather than synthetic filler.

Sources (24 cited facts, 0 third-party)

Each fact links to the page it came from; the badge says who published it (how to read sources).