TL;DR
DeepSeek V4 Flash cost $6.23 per 1,000 next-action decisions in the August 8, 2026 study, with 98.1% valid structured output and a judge mean of 75.3. Its gap with Luna narrowed sharply under another judge. On the separate ship-traffic research task, all six reviews were D, with a reported $5.17 run cost. Low decision cost did not establish reliable completed research.
- Exact tested ID
- deepseek/deepseek-v4-flash
- Published studies
- 3
- Provider
- deepseek
Planning models on 29 frozen conversations: compare the tested outputs
Tick: recorded mean score. Whisker: mean ± one recorded standard error (SE), not a confidence interval. Overlap alone does not establish a ranking.
16-model campaign, 29 frozen conversations; 14 model rows shown in the public table. Published 2026-08-08. Values retain their original units and grading scale.
Show the table · 2 rows, 5 columns
Scroll sideways to compare all columns. Select a heading to sort.
| deepseek/deepseek-v4-flash | 0.493 | 0.080 | $0.0032 | 12.4s |
| meta/muse-spark-1.1 | 0.759 | 0.079 | $0.0229 | 9.6s |
No matching rows. Clear the filter to see all records.
16-model campaign, 29 frozen conversations; 14 model rows shown in the public table. Published 2026-08-08. Values retain their original units and grading scale.
deepseek/deepseek-v4-flash was tested on 2026-08-08. 16-model campaign, 29 frozen conversations; 14 model rows shown in the public table. Its row and the other published arm use the same task and measurement units.
Next-action execution model bakeoff: compare the tested outputs
These exact August 8, 2026 execution-table values compare DeepSeek V4 Flash with the displayed Luna Pro arm. Labels are rounded; the full table preserves valid-output rates and judge scores. This is historical billing.
Show the table · 2 rows, 5 columns
Scroll sideways to compare all columns. Select a heading to sort.
| deepseek/deepseek-v4-flash | 75.3 | 98.1% | $6.23 | 17.9s |
| openai/gpt-5.6-luna-pro | 91.7 | 100.0% | $25.05 | 9.9s |
No matching rows. Clear the filter to see all records.
23-model campaign; 69 frozen decisions × 3 samples; 12 published model rows. Published 2026-08-08. Values retain their original units and grading scale.
deepseek/deepseek-v4-flash was tested on 2026-08-08. 23-model campaign; 69 frozen decisions × 3 samples; 12 published model rows. Its row and the other published arm below use the same task and measurement units.
Five models completing one sandbox research task: compare the tested outputs
5 models on the same single research task; real Modal virtual machines. Published 2026-08-08. Values retain their original units and grading scale.
Show the table · 2 rows, 4 columns
Scroll sideways to compare all columns. Select a heading to sort.
| deepseek/deepseek-v4-flash | D → D → D → D → D → D | 181m | $5.17 |
| openai/gpt-5.6-luna | D → B+ → D → B+ → D → B+ | 59m | $2.63 |
No matching rows. Clear the filter to see all records.
5 models on the same single research task; real Modal virtual machines. Published 2026-08-08. Values retain their original units and grading scale.
deepseek/deepseek-v4-flash was tested on 2026-08-08. 5 models on the same single research task; real Modal virtual machines. Its row and the other published arm below use the same task and measurement units.
What this version helps you choose
Luna had better recorded operating results on this agent workload. Keep the sharply reduced execution score gap under a different judge beside the choice.
Route a request and plan the work: follow the published task
A useful planner chooses the right kind of work before doing it. The complete customer conversations and system prompts remain private.
Published task summary
An incoming request and its existing conversation.
Choose whether to write a plan, call a catalog tool, ask for clarification or continue an existing agent.
Read what came back
A routing decision and plan. Every model receives the same saved conversation; the test compares that decision before fresh research begins.
Replay 29 frozen conversations across the model arms, then compare graded plan scores, billing and response time.
Planning models on 29 frozen conversations
Inspect the original single-model metrics
For routing and planning, Deepseek V4 Flash recorded a mean score of 0.493 with standard error 0.080. Historical billing was $0.0032 per decision, and median response time was 12.4s. Compare that cost and response time with the other models in the same 29-conversation study; the mean alone does not establish a decisive quality advantage.
| Published metric | Value |
|---|---|
| Score | 0.493 |
| +/- SE | 0.080 |
| $ / decision | $0.0032 |
| p50 | 12.4s |
No matching rows. Clear the filter to see all records.
Planning models on 29 frozen conversations: read these measurements
Read the complete rubric, sample and billing notes
16-model campaign, 29 frozen conversations; 14 model rows shown in the public table
Route an incoming request: write a plan, call a catalog tool, ask for clarification or continue an existing agent.
Scoring
Top planning means were close at this sample size. The source's paired comparison with Opus was inconclusive.
Cost units
Cost is provider-billed dollars per decision. Prefix reuse can make it differ from list price multiplied by every input token.
Time units
p50 is median latency in seconds, not the duration of complete research.
Frozen conversation
A saved input kept identical across model arms so their decisions can be compared.
Standard error (SE)
An estimate of uncertainty around the sample mean. A small mean difference needs more evidence than the ranking alone.
Next-action execution model bakeoff
Inspect the original single-model metrics
For the next allowed action, Deepseek V4 Flash recorded a judge mean of 75.3 and produced valid structured output 98.1% of the time. Historical billing was $6.23 per 1,000 decisions; production median response time was 17.9s. Decision quality, readable output and operating cost answer different questions.
| Published metric | Value |
|---|---|
| Mean score | 75.3 |
| Schema valid | 98.1% |
| $ / 1k decisions | $6.23 |
| Prod p50 | 17.9s |
No matching rows. Clear the filter to see all records.
Choose the next allowed agent action: follow the published task
A well-formed action can still be the wrong decision. Keep output validity and judged decision quality separate.
Published task summary
The existing plan, conversation and results returned by tools.
Choose one next allowed action, or finish with an answer when the evidence is sufficient.
Read what came back
A structured action. The replay records the proposed action without executing its tools.
Replay 69 saved decisions three times each. Check the required output structure separately from the judge score.
Next-action execution model bakeoff: read these measurements
Read the complete rubric, sample and billing notes
23-model campaign; 69 frozen decisions × 3 samples; 12 published model rows
Choose one next action in an unfinished trading-agent workflow. This replay measures individual decisions, not a fresh complete trading run.
Scoring
Mean action score comes from the judge. Schema validity is the share of actions matching the required output structure; a valid structure can still contain the wrong action. Switching judges reduced the Luna-versus-DeepSeek score gap from 13.9 to 1.5 points.
Cost units
Cost uses historical provider billing per 1,000 decisions. Reused prefixes earned cache discounts; the cached-token split was not retained.
Time units
Prod p50 is median production response latency, separate from replay quality and complete agent wall time.
Five models completing one sandbox research task
Inspect the original single-model metrics
On the ship-traffic research task, Deepseek V4 Flash returned the grade sequence D → D → D → D → D → D over 181m, with reported run cost $5.17. Read the whole sequence: a favorable review followed by a failure exposes a repair that did not hold. A missing cost remains missing, and uneven CPU throttling limits timing comparisons.
| Published metric | Value |
|---|---|
| Grade trajectory | D → D → D → D → D → D |
| Wall | 181m |
| Cost | $5.17 |
No matching rows. Clear the filter to see all records.
Write research code and repair its artifacts: follow the published task
The grade sequence shows whether a repair held up at the next review. One B+ followed by D is a different outcome from repeated favorable reviews.
Run one concrete research task
Correlate Strait of Hormuz ship traffic with the next USO trading session.
Write and run research code on real Modal machines, return the artifacts for review and repair defects raised by the grader.
Inspect the recorded repairs and regressions
Luna returned D → B+ → D → B+ → D → B+; GLM returned D → D → B+ → D → D → D. DeepSeek returned D on all six reviews.
The source found self-check code that assigned the same variable to delivered_stats and recomputed, then hard-coded a match. That explains why a passing self-check could survive while the delivered analysis still failed review. The publication does not assign this particular code defect to one model in the table.
Five models completing one sandbox research task: read these measurements
Read the complete rubric, sample and billing notes
5 models on the same single research task; real Modal virtual machines
Correlate Strait of Hormuz ship traffic with the next USO trading session.
Scoring
Grades describe the delivered artifact after successive reviews. Oscillation exposed holes in agent-written self-checks; a successful process exit was insufficient.
Cost units
Cost is historical billed run cost. Flash Lite has no reported total after failing out; a dash is missing data, not zero.
Time units
Wall time covers the complete workflow. Uneven web-tier CPU throttling affected timing, so the source treats it as a soft comparison.
Sandbox
An isolated machine where the agent runs research code.
No comments yet.