← Back to Cross-Model Leaderboard
MODEL EVALUATION DASHBOARD

DeepSeek V4 Flash

Evaluating agent task completion, skill acceleration, latency, and token economics on 31 benchmark tasks.
Model: openrouter/deepseek/deepseek-v4-flash-0731 • Sampling: k=1 (62 runs total)

🌟 With Skills
87.1%
PASS RATE (27/31)
Pass rate achieved when the agent utilizes AtomisticSkills domain tools and workflows.
52m
MEAN LATENCY
Mean wall-clock execution time for the agent to complete the task with domain skills.
60.5k
MEAN OUT TOKENS
Mean output tokens generated by the LLM (total completion tokens, including reasoning/thinking tokens where applicable) when utilizing domain skills.
89.2%
CACHE HIT RATE
Percentage of input tokens served from provider KV cache during with-skills runs (Cached Tokens / Total Input Tokens).
⚖️ No Skills
74.2%
PASS RATE (23/31)
Baseline pass rate achieved when the agent attempts tasks using only general coding tools without domain skills.
65m
MEAN LATENCY
Mean wall-clock execution time for the agent without domain skills.
78.0k
MEAN OUT TOKENS
Mean output tokens generated without domain skills (total completion tokens, including reasoning/thinking tokens). Without structured skills, models frequently burn more tokens exploring.
95.9%
CACHE HIT RATE
Percentage of input tokens served from provider KV cache during baseline no-skills runs.
+12.9%
SKILL BOOST Δ
Absolute pass rate gain achieved over the no-skills baseline: (With-Skills Pass) - (No-Skills Pass).
$10.92
TOTAL COST (62 RUNS)
Total LLM inference cost computed from provider token pricing, incorporating cached input discounts.
Discipline Task Name With Skills No Skills All Solvers Time (With / No) Out Tokens (With / No) Action
chemistry bde-weak-bonds 1/1 1/1 2/2 (100.0%) 77m/67m 58.6k/68.7k View Details →
chemistry conformer-boltzmann-ranking 1/1 0/1 1/2 (50.0%) 14m/51m 33.7k/28.3k View Details →
chemistry gas-thermochemistry-equilibrium 1/1 1/1 2/2 (100.0%) 14m/473s 23.9k/40.7k View Details →
chemistry ghs-hazard-summary 1/1 0/1 1/2 (50.0%) 319s/50m 7.3k/31.4k View Details →
chemistry ir-spectrum-match 1/1 1/1 2/2 (100.0%) 20m/484s 20.8k/5.6k View Details →
chemistry mof-dac-screening 1/1 1/1 2/2 (100.0%) 24m/106m 126.6k/385.0k View Details →
chemistry nmr-reaction-kinetics 1/1 1/1 2/2 (100.0%) 453s/32m 15.8k/56.2k View Details →
drug-discovery admet-lead-triage 1/1 1/1 2/2 (100.0%) 11m/29m 28.9k/19.6k View Details →
drug-discovery docking-microstate-enrichment 0/1 1/1 1/2 (50.0%) 24m/23m 75.5k/174.4k View Details →
drug-discovery ecfp-analog-diversity 1/1 1/1 2/2 (100.0%) 14m/14m 14.2k/9.5k View Details →
drug-discovery ligand-box-definition 1/1 1/1 2/2 (100.0%) 276s/45m 14.1k/30.2k View Details →
drug-discovery ligand-md-stability 1/1 1/1 2/2 (100.0%) 21m/54m 26.6k/36.2k View Details →
drug-discovery pose-quality-filter 1/1 1/1 2/2 (100.0%) 450s/267s 10.6k/27.7k View Details →
drug-discovery symmetry-redocking-rmsd 1/1 1/1 2/2 (100.0%) 382s/23m 13.6k/24.2k View Details →
machine-learning committee-uncertainty-flagging 1/1 1/1 2/2 (100.0%) 429s/444s 7.7k/4.3k View Details →
machine-learning mlip-error-benchmark 1/1 1/1 2/2 (100.0%) 311s/210s 10.6k/20.1k View Details →
materials-science camgsi-elasticity-vrh 1/1 1/1 2/2 (100.0%) 110m/66m 166.4k/119.2k View Details →
materials-science co-adsorption-cu111 1/1 1/1 2/2 (100.0%) 43m/26m 51.9k/56.3k View Details →
materials-science convex-hull-stability 0/1 0/1 0/2 (0.0%) 145m/32m 154.6k/46.8k View Details →
materials-science crconi-order-disorder 0/1 1/1 1/2 (50.0%) 356m/531m 349.5k/655.1k View Details →
materials-science electrochemical-window 0/1 1/1 1/2 (50.0%) 17m/82m 14.9k/106.7k View Details →
materials-science li-qha-simulation 1/1 0/1 1/2 (50.0%) 117m/45m 118.7k/65.9k View Details →
materials-science lifepo4-intercalation-voltage 1/1 0/1 1/2 (50.0%) 93m/518s 82.7k/12.9k View Details →
materials-science mgo-vacancy-energy 1/1 1/1 2/2 (100.0%) 85m/35m 112.0k/39.5k View Details →
materials-science ni3al-surface-energy 1/1 1/1 2/2 (100.0%) 196m/58m 160.9k/51.1k View Details →
materials-science nist-janaf-query 1/1 1/1 2/2 (100.0%) 504s/258s 3.1k/4.2k View Details →
materials-science polymorph-free-energy 1/1 0/1 1/2 (50.0%) 80m/434m 45.7k/50.0k View Details →
materials-science si-eos-fit 1/1 1/1 2/2 (100.0%) 30m/26m 34.0k/28.2k View Details →
materials-science sic-thermal-conductivity 1/1 0/1 1/2 (50.0%) 34m/65m 43.6k/86.2k View Details →
materials-science synthesis-precursor-recommendation 1/1 1/1 2/2 (100.0%) 10m/34m 4.5k/66.6k View Details →
materials-science xrd-mixture-phase-fit 1/1 0/1 1/2 (50.0%) 34m/44m 44.4k/68.2k View Details →