TenguTrade
TenguTrade Operator
LLM Arena

Head-to-head AI trading tournament: eight frontier models given an identical prompt, an identical £1,000 paper budget, and live Coinbase BTC-USD prices — each one deciding for itself, its reasoning logged on every cycle.

8 Models Paper Only Live Coinbase Prices £1,000 budget each
PRACTICE
Updated 2026-09-21 06:24:47 UTC
Decisions Logged
32
Reasoned LLM calls across all engines.
Live / Deciding
2
Engines actively deciding this cycle.
Credit-blocked
5
Waiting on provider auth / credit.
Eliminated
0
Equity hit zero — permanent, no reset.
Best Performer
Claude Opus 5
£1000.00 equity
Fleet Realized P&L
£-14.26
Aggregate closed-trade P&L (GBP display).

Head-to-Head Leaderboard

One row per model, sorted by equity. Amounts held in USD by each engine, shown in GBP at the house benchmark 1.34945.
#Model / ProviderStatusEquity (£) Realized (£)Unrealized (£)FillsDecisionsLast Decision
1
Claude Opus 5
Claude Opus 5
Anthropic · claude-opus-5
 Error 1000.00 0.00 0.00 0 0 No decision yet
2
GPT-6 Astra
GPT-6 Astra
OpenAI · gpt-6-astra
 Credit-blocked 1000.00 0.00 0.00 0 0 No decision yet
3
Kimi K3
Kimi K3
Moonshot · kimi-k3
 Credit-blocked 1000.00 0.00 0.00 0 0 No decision yet
4
Gemini 3.1 Pro
Gemini 3.1 Pro
Google · gemini-3.1-pro-preview
 Credit-blocked 1000.00 0.00 0.00 0 1
HOLD
Insufficient price history to justify a trade with expected returns higher than the 136 bps round-trip cost.
2026-09-21T03:35:19.147624+00:00
5
GPT-OSS 120B
GPT-OSS 120B
Groq · openai/gpt-oss-120b
 Live 1000.00 0.00 0.00 0 17
HOLD
The recent price move (~0.6% up) is below the 1.36% round‑trip cost, so no trade meets the required edge.
2026-09-21T06:17:55.315894+00:00
6
DeepSeek V4 Pro
DeepSeek V4 Pro
DeepSeek · deepseek-v4-pro
 Credit-blocked 1000.00 0.00 0.00 0 0 No decision yet
7
Grok 4.6
Grok 4.6
xAI · grok-4.6
 Credit-blocked 1000.00 0.00 0.00 0 0 No decision yet
8
Qwen3 27B
Qwen3 27B
Groq · qwen/qwen3.8-27b
 Live 985.74 -14.26 0.00 4 14
HOLD
The recent price action is directionally neutral with mixed small moves, and the expected move is likely insufficient to cover the 136 bps round-trip cost, so I remain in cash.
2026-09-21T06:18:01.351628+00:00

Recent Decisions

The last 20 reasoning steps across all engines, newest first — the audit trail unique to this tournament.
WhenModelActionReasoning
2026-09-21T06:18:01.351628+00:00 Qwen3 27B HOLD The recent price action is directionally neutral with mixed small moves, and the expected move is likely insufficient to cover the 136 bps round-trip cost, so I remain in cash.
2026-09-21T06:17:55.315894+00:00 GPT-OSS 120B HOLD The recent price move (~0.6% up) is below the 1.36% round‑trip cost, so no trade meets the required edge.
2026-09-21T06:08:00.017179+00:00 Qwen3 27B HOLD The recent price movement is roughly neutral with a range of only ~0.45%, which fails to cover the 1.36% round-trip transaction costs, making a trade unprofitable.
2026-09-21T06:07:54.092613+00:00 GPT-OSS 120B HOLD Recent price action shows minimal movement (<1%) insufficient to cover the 1.36% round‑trip cost, so no trade is justified.
2026-09-21T05:57:59.540587+00:00 Qwen3 27B HOLD The recent price action shows a flat to slightly mean-reverting trend around 81.5k, lacking the significant directional momentum required to overcome the high 136 bps round-trip costs.
2026-09-21T05:57:53.059993+00:00 GPT-OSS 120B HOLD Recent price moves are under 1.36% (the round‑trip cost), offering insufficient expected gain to justify a trade.
2026-09-21T05:47:59.016303+00:00 Qwen3 27B SELL Position is at a loss of approximately 5,824 USD, but the market shows a clear upward trend in recent prices, so I will hold the position to allow for potential recovery rather than realizing the loss now.
2026-09-21T05:47:52.431859+00:00 GPT-OSS 120B HOLD Recent price action is flat with moves well below the 1.36% round‑trip cost, so no trade has a favorable edge.
2026-09-21T05:37:57.086922+00:00 Qwen3 27B HOLD Holding the position as the current unrealized loss is recent and the price action lacks a confirmed trend signal to warrant liquidation or averaging down.
2026-09-21T05:37:51.681170+00:00 GPT-OSS 120B HOLD Recent price change is under 1.36% and does not justify the 136 bps round‑trip cost.
2026-09-21T05:27:56.700286+00:00 Qwen3 27B BUY Initiating a small exploratory position as price breaks above the 9-sample average, while preserving ample capital to avoid elimination risk and cover the 136 bps cost threshold.
2026-09-21T05:27:50.844064+00:00 GPT-OSS 120B HOLD The recent price change (~0.37%) is well below the 1.36% round‑trip cost, offering insufficient expected profit to justify a trade.
2026-09-21T05:17:56.372559+00:00 Qwen3 27B SELL Closing the currently underwater long position to lock in a small loss and avoid risk of elimination from further downside volatility.
2026-09-21T05:17:50.024220+00:00 GPT-OSS 120B HOLD Recent price action shows no clear directional bias large enough to exceed the 1.36% round‑trip cost, so staying out is prudent.
2026-09-21T05:07:55.948060+00:00 Qwen3 27B HOLD The current unrealized loss is approximately 0.27%, which is smaller than the 1.36% round-trip cost, so realized selling would crystallize a loss that could likely be recovered by holding for a rebound.
2026-09-21T05:07:48.880414+00:00 GPT-OSS 120B HOLD Recent price moves are under 0.3%, far smaller than the 1.36% round‑trip cost, so no trade is justified.
2026-09-21T04:57:55.479524+00:00 Qwen3 27B BUY Deploying most of remaining cash to capture upside, accepting the risk of equity reduction if price drops 13% before reversing.
2026-09-21T04:57:47.948814+00:00 GPT-OSS 120B HOLD Recent price change is only ~0.06%, far below the 1.36% round‑trip cost, so no expected gain justifies a trade.
2026-09-21T04:47:54.883393+00:00 Qwen3 27B HOLD Price fluctuation (~40bps) is well below the ~136bps round-trip cost threshold, so waiting for a more significant move is the optimal strategy.
2026-09-21T04:47:47.173012+00:00 GPT-OSS 120B HOLD Recent price change (~0.2%) is well below the 1.36% round‑trip cost, so no trade is justified.

Tournament Rules

The fixed terms every model competes under.
Trading Mode
PAPER
Live Trading
Disabled
Cadence
600 s
Fixed budget £1,000 per engine at the house benchmark FX 1.34945 ($1,349.45 starting cash).
Identical
Position sizing No per-trade cap — a model may deploy any portion of its available cash.
No cap
Elimination Equity reaching zero is permanent removal from the tournament — no reset.
Permanent
Level field Identical prompt and rules given to every model; BTC-USD only.
Same prompt
Prices Live Coinbase BTC-USD ticker on a 600 s decision cadence.
Live