FrissoncraftCS2 reasoning tokens
DUST II · OPENROUTER · NOVEMBER 11, 2026

CS2 reasoning tokens

72 full traces. One question.

Give me the routes + actions of each player.

Full traces
72
Models
3
Situations
5
Repetitions
5 per pair

Three models received five fixed situations from a professional Dust II recording, five times each. Every returned response is here, including the full thinking text, exact prompt, player routes/actions and timing.

75 calls produced 72 responses. One timeout and two provider rate limits produced no trace. Two answers vary from the requested format and are included unchanged. No CS2 simulation was run.

Download all 72 full traces ↓Download timing CSV ↓Publication manifest ↗

Request timing

End-to-end latency for returned responses, including provider waiting and generation.

ModelResponsesMedian timeRange
DeepSeek V4 Pro 081322/25485.88 s100.67–923.22 s
Qwen3.8 2.4T A95B25/2564.04 s44.40–89.19 s
Kimi K3.1 Pro25/2523.64 s11.31–103.69 s

All traces

72 traces

SituationModelRunTimeThinking tokensOutput formatTrace
Long approachDeepSeek V4 Pro 08133/5452.19 s6674RecordedRead full trace ↗
Long approachDeepSeek V4 Pro 08134/5375.30 s5748RecordedRead full trace ↗
Long approachDeepSeek V4 Pro 08135/5467.30 s7310RecordedRead full trace ↗
Long approachQwen3.8 2.4T A95B1/580.71 s2386RecordedRead full trace ↗
Long approachQwen3.8 2.4T A95B2/570.34 s1915RecordedRead full trace ↗
Long approachQwen3.8 2.4T A95B3/554.17 s1442RecordedRead full trace ↗
Long approachQwen3.8 2.4T A95B4/549.00 s1319RecordedRead full trace ↗
Long approachQwen3.8 2.4T A95B5/561.48 s1955RecordedRead full trace ↗
Long approachKimi K3.1 Pro1/523.31 s2263RecordedRead full trace ↗
Long approachKimi K3.1 Pro2/522.03 s2176RecordedRead full trace ↗
Long approachKimi K3.1 Pro3/527.14 s2644RecordedRead full trace ↗
Long approachKimi K3.1 Pro4/535.40 s3516Format variationRead full trace ↗
Long approachKimi K3.1 Pro5/543.21 s2135RecordedRead full trace ↗
Mid-round disadvantageDeepSeek V4 Pro 08131/5683.53 s10617RecordedRead full trace ↗
Mid-round disadvantageDeepSeek V4 Pro 08132/5749.70 s12128RecordedRead full trace ↗
Mid-round disadvantageDeepSeek V4 Pro 08133/5599.16 s9890RecordedRead full trace ↗
Mid-round disadvantageDeepSeek V4 Pro 08134/5923.22 s13576RecordedRead full trace ↗
Mid-round disadvantageDeepSeek V4 Pro 08135/5703.68 s10496RecordedRead full trace ↗
Mid-round disadvantageQwen3.8 2.4T A95B1/561.60 s1377RecordedRead full trace ↗
Mid-round disadvantageQwen3.8 2.4T A95B2/582.24 s2255RecordedRead full trace ↗
Mid-round disadvantageQwen3.8 2.4T A95B3/589.19 s2632RecordedRead full trace ↗
Mid-round disadvantageQwen3.8 2.4T A95B4/575.88 s2390RecordedRead full trace ↗
Mid-round disadvantageQwen3.8 2.4T A95B5/567.47 s2095RecordedRead full trace ↗
Mid-round disadvantageKimi K3.1 Pro1/526.06 s3068RecordedRead full trace ↗
Mid-round disadvantageKimi K3.1 Pro2/518.49 s1784RecordedRead full trace ↗
Mid-round disadvantageKimi K3.1 Pro3/523.64 s1739RecordedRead full trace ↗
Mid-round disadvantageKimi K3.1 Pro4/539.86 s1687RecordedRead full trace ↗
Mid-round disadvantageKimi K3.1 Pro5/522.23 s2347RecordedRead full trace ↗
B post-plantDeepSeek V4 Pro 08131/5249.35 s3495RecordedRead full trace ↗
B post-plantDeepSeek V4 Pro 08132/5100.67 s1348RecordedRead full trace ↗
B post-plantDeepSeek V4 Pro 08133/5153.79 s2034RecordedRead full trace ↗
B post-plantDeepSeek V4 Pro 08134/5205.42 s2981RecordedRead full trace ↗
B post-plantDeepSeek V4 Pro 08135/5162.98 s2524RecordedRead full trace ↗
B post-plantQwen3.8 2.4T A95B1/567.24 s1753RecordedRead full trace ↗
B post-plantQwen3.8 2.4T A95B2/564.47 s1430RecordedRead full trace ↗
B post-plantQwen3.8 2.4T A95B3/572.94 s1501RecordedRead full trace ↗
B post-plantQwen3.8 2.4T A95B4/558.36 s1504RecordedRead full trace ↗
B post-plantQwen3.8 2.4T A95B5/561.67 s1565RecordedRead full trace ↗
B post-plantKimi K3.1 Pro1/536.72 s546RecordedRead full trace ↗
B post-plantKimi K3.1 Pro2/514.65 s1470RecordedRead full trace ↗
B post-plantKimi K3.1 Pro3/515.74 s983RecordedRead full trace ↗
B post-plantKimi K3.1 Pro4/515.98 s1562RecordedRead full trace ↗
B post-plantKimi K3.1 Pro5/511.31 s971RecordedRead full trace ↗
Protect the advantageDeepSeek V4 Pro 08131/5515.87 s7678RecordedRead full trace ↗
Protect the advantageDeepSeek V4 Pro 08132/5523.44 s7492Format variationRead full trace ↗
Protect the advantageDeepSeek V4 Pro 08133/5690.21 s10564RecordedRead full trace ↗
Protect the advantageDeepSeek V4 Pro 08134/5433.61 s6309RecordedRead full trace ↗
Protect the advantageDeepSeek V4 Pro 08135/5506.90 s7524RecordedRead full trace ↗
Protect the advantageQwen3.8 2.4T A95B1/564.04 s1701RecordedRead full trace ↗
Protect the advantageQwen3.8 2.4T A95B2/581.20 s2195RecordedRead full trace ↗
Protect the advantageQwen3.8 2.4T A95B3/556.53 s1261RecordedRead full trace ↗
Protect the advantageQwen3.8 2.4T A95B4/564.17 s1918RecordedRead full trace ↗
Protect the advantageQwen3.8 2.4T A95B5/555.72 s1554RecordedRead full trace ↗
Protect the advantageKimi K3.1 Pro1/533.87 s3162RecordedRead full trace ↗
Protect the advantageKimi K3.1 Pro2/525.80 s3070RecordedRead full trace ↗
Protect the advantageKimi K3.1 Pro3/5103.69 s1890RecordedRead full trace ↗
Protect the advantageKimi K3.1 Pro4/537.86 s3159RecordedRead full trace ↗
Protect the advantageKimi K3.1 Pro5/529.17 s2514RecordedRead full trace ↗
Late commitmentDeepSeek V4 Pro 08131/5504.46 s8204RecordedRead full trace ↗
Late commitmentDeepSeek V4 Pro 08132/5373.34 s5615RecordedRead full trace ↗
Late commitmentDeepSeek V4 Pro 08133/5439.76 s6878RecordedRead full trace ↗
Late commitmentDeepSeek V4 Pro 08134/5613.69 s9495RecordedRead full trace ↗
Late commitmentQwen3.8 2.4T A95B1/566.39 s1565RecordedRead full trace ↗
Late commitmentQwen3.8 2.4T A95B2/558.05 s1432RecordedRead full trace ↗
Late commitmentQwen3.8 2.4T A95B3/562.98 s1859RecordedRead full trace ↗
Late commitmentQwen3.8 2.4T A95B4/561.72 s1639RecordedRead full trace ↗
Late commitmentQwen3.8 2.4T A95B5/544.40 s1098RecordedRead full trace ↗
Late commitmentKimi K3.1 Pro1/528.76 s1826RecordedRead full trace ↗
Late commitmentKimi K3.1 Pro2/521.69 s2167RecordedRead full trace ↗
Late commitmentKimi K3.1 Pro3/522.39 s1735RecordedRead full trace ↗
Late commitmentKimi K3.1 Pro4/519.62 s1724RecordedRead full trace ↗
Late commitmentKimi K3.1 Pro5/520.75 s2089RecordedRead full trace ↗

Collection notes

Only the Terrorists’ combined observations were supplied. Enemy sightings retain their age. Match names, later outcomes and future movements were excluded from prompts. The returned thinking is provider-exposed text, not a guarantee that all internal reasoning is visible.

DeepSeek and Kimi used high reasoning effort; Qwen used medium because that endpoint does not support high. The requested temperature was 0.7, with a 16,384-token output limit; provider thinking modes may override sampling settings. The first 14 attempts used three workers, then the remaining 61 used nine. There were no retries or follow-up model calls.