RAM labels describe the test computer, not this model’s memory requirement. Speed comparisons keep the 32 GB and 128 GB runs separate.
Runtime build: unrecorded · Local · test machine with 32 GB RAM · concurrency: 1
Thinking status: verified · probe 1 · scope Saved profile-level probe; exact historical template/build scope is not fully recorded..
Requested off does not prove an absence of reasoning. Saved checks do not establish unrecorded historical template or build details.
Overall across tasks
33.8 / 100
Nine categories, equally weighted · 9/9 eligible
Complete coverage for this setup. Quality only; speed is separate. How the index works →
What does it do well?
Select a category to explore the answers. Bars show what happened on every planned task. Incorrect answers and failures to finish both earn zero; technical problems and untested tasks stay unscored.
Exact scores and task times by suite
Each suite has its own metric and fixed task bank. The balanced index weights categories equally; individual scores and evidence remain available below.
This analysis covers this configuration across its recorded thinking conditions and test suites.
Use Gemma 4 E2B Q4_0 with thinking for small-project assistance, calculations from passages and keeping answers aligned with changing sources. It handles most small coding tasks here, but that does not carry over to larger tool-driven jobs. Research-paper question answering is a particularly poor fit. Puzzle solving also performed well in the tested thinking-off setup.
Original analysis generated by GPT Astra · · Edited by Codex; agent reviewed by Codex · publication authorized by user
About this summary
Codex reviewed this assessment against a newer evidence edition. The original analyst, generation date and source draft remain recorded separately from the editorial revision.
Analyst provider
openai
Generation route
Codex CLI · codex-cli 0.162.1 · ChatGPT account
Requested model
gpt-6-astra
Returned model
Not supplied; identity is recorded from the request only.
Exact provider snapshot
Exact provider snapshot not supplied.
Reasoning effort
high
Prompt version
model-summary-reader-v6
Generation time
2026-10-12T05:24:18.915418Z
Review time
2026-10-22T07:26:27.148335+00:00
Evidence reviewer
Codex (agent)
Publication authorization
User · 2026-10-22T05:55:43.625712+00:00 · Go ahead and complete the remaining work. Spin up agents: luna, terra, sol, astra, as needed to get it done. Narrative summaries: these are the astra generated summaries? Go ahead and get those completed as well. Once all of this is done, you may push the completed result.
Handles calculations from passages and keeps answers aligned with evidence. With thinking enabled, it passed every case in these small tests of passage calculations, resisting incorrect challenges, and updating answers when sources changed.
Logic and number puzzles are a useful application: with thinking enabled, it fully solved most of these puzzles. That supports using it as a reasoning assistant, while still checking its answers.
Research-paper question answering is extremely weak in either mode. Answer matching remains very low even with thinking, so its success with short passages should not be treated as evidence of useful paper analysis.
Larger practical jobs expose a substantial limit: even with thinking, it completed only a small minority of these tasks. Do not extrapolate its small-project usefulness to autonomous app building or longer investigations.
Keep thinking on for small-project work and passage calculations. Paired tests showed substantial accuracy gains for a modest increase in average task time—a worthwhile tradeoff for these everyday uses.
Thinking is optional for puzzles: it produced a modest score improvement with little extra average time, but some answers worsened. Enable it for the aggregate gain, not a guarantee of better answers.
Evidence for this judgment
Logic and number puzzles · Thinking off versus on
Thinking off → on
58.9/100 · Mean puzzle credit → 64.7/100 · Mean puzzle credit
With thinking enabled, it recorded narrow leads over the closest tested small dense alternative on puzzle scores and selected hard problems. Research-paper answer matching, however, trails the other tested small dense models substantially.
Evidence for this judgment
Logic and number puzzles Thinking on. Small · under 5B · dense. Rank 1 of 5 in this recorded comparison.
Published hard problems Thinking on. Small · under 5B · dense. Rank 1 of 5 in this recorded comparison.
Reading research papers Thinking on. Small · under 5B · dense. Rank 5 of 5 in this recorded comparison.
Other mode, this model only: thinking off · 58.9/100 · Mean puzzle credit.
How this comparison works
Average credit per task; partly correct results can earn partial credit. Family comparisons show the labeled Instruct and reasoning versions. Other peer comparisons use the same requested thinking setting, not necessarily each model’s best mode. Small differences on these short tests may not hold on other tasks. Peer membership varies by test coverage and compatible settings.
Comparison saved with the 2026-10-12 assessment. Task-specific results.
Published hard problemsThinking on · Small · under 5B · dense
Answer selected difficult math and reasoning questions from LiveBench.
Mean answer credit · 0–100 points · higher is better 11 of 11 tasks scored for each model.
Gaps are relative to this model. Treat small gaps cautiously.
Other mode, this model only: thinking off · 20.3/100 · Mean answer credit.
How this comparison works
Average credit per task; partly correct results can earn partial credit. Family comparisons show the labeled Instruct and reasoning versions. Other peer comparisons use the same requested thinking setting, not necessarily each model’s best mode. Small differences on these short tests may not hold on other tasks. Peer membership varies by test coverage and compatible settings.
Comparison saved with the 2026-10-12 assessment. Task-specific results.
Reading research papersThinking on · Small · under 5B · dense
Answer a question about a paper and identify the paragraphs supporting the answer.
Answer token F1 · 0–100 points · higher is better 8 of 8 tasks scored for each model.
Gaps are relative to this model. Treat small gaps cautiously.
Other mode, this model only: thinking off · 0.0/100 · Answer token F1.
How this comparison works
Answer overlap with the reference answer; partial matches earn credit. Family comparisons show the labeled Instruct and reasoning versions. Other peer comparisons use the same requested thinking setting, not necessarily each model’s best mode. Small differences on these short tests may not hold on other tasks. Peer membership varies by test coverage and compatible settings.
Comparison saved with the 2026-10-12 assessment. Task-specific results.
With thinking off, puzzle credit exceeded both members of the matched speed comparison; its time was essentially tied with Ministral 3 14B and shorter than Qwen3.5 2B Q4_K_M.
Evidence for this judgment
Speed: Logic and number puzzles Thinking off. Matching test-machine label and execution settings; spans model sizes. Timings include unsuccessful tasks, so read them alongside quality.
Speed and task success: Logic and number puzzlesThinking off · Matching test-machine label and settings · Across model sizes
Shorter task time is better only when the result is useful. These times include failures, tools and grading. Token generation speed measures a different part of the work.
Results from this test setup, not a promise for your computer. Hardware, software, tokenizers and answer length can change the ordering. Historical runtime builds were not always captured. Comparison saved with the 2026-10-12 assessment.
Test machine: AMD Ryzen 5 5600X (6 cores / 12 threads) · AMD Radeon RX 6700 XT (12 GB VRAM) · 32 GB or 128 GB system RAM, never mixed in one timing group. Full test machine details
Worth knowing
Some examples cannot be inspected fully because the publication export contains redactions, source links instead of paper text, and withheld project artifacts.
Evidence for this judgment
Source limitation Some selected evidence has recorded publication redactions.
Thinking on took 5.8 seconds more per matched timed question, on average.
All matched quality and timing measurements
Matched measurement
Recorded result
Average credit, off → on
31.3% → 81.3%
Mean paired credit change
+50.0 percentage points
Question outcomes
8 improved · 8 unchanged · 0 worse
Mean added task time
+5.8s
Mean task time, off → on
3.7s → 9.5s
Median task time, off → on
2.2s → 8.1s
Median of paired time changes
+5.5s
Total added task time
+93.6s
Ratio of summed times, on / off
2.6×
Quality and timing can cover different subsets. Median paired change is calculated per question; it is not the difference between the two medians. Task time includes tools and grading.
The user asks where the project's retry-limit setting is defined and what value it holds. This exercises locating a named setting in code and reporting both its location and value.
The user asks for the release date mentioned in a project notes document. This exercises reading a specified file and extracting a particular fact rather than guessing from outside knowledge.
The user asks to change the project's retry limit to a specified value. This exercises finding the relevant setting and making a focused code edit rather than merely describing the change.
The user asks to create a text file containing exactly one supplied line. This exercises basic file creation and precise instruction following, including avoiding extra text in the file.
The user asks what the most recent Git commit changed in the project. This exercises inspecting recorded changes and turning them into a clear explanation of what was modified.
The user asks how many Python files are in the application folder and which is largest. This exercises identifying relevant files, counting them, and comparing their sizes.
The user asks to add a function that converts byte counts into readable sizes and remove the related TODO comment. This exercises implementing a small utility and updating the surrounding code consistently.
The user asks to find every TODO comment in the Python code and append each to the project's task list on a separate line. This combines searching across files with collecting text into an existing document.
The user asks to investigate a failing test and fix the application code without changing the test. This exercises diagnosing a bug while respecting an explicit limit on what may be edited.
The user asks to rename a network function and update every place in the project that uses it. This exercises tracking references across files and making a coordinated change rather than editing only the definition.
The user asks to find a proxy configuration file somewhere under the project's parent folder and report its host and port. This exercises searching beyond the project folder and reading configuration values.
The user asks which error code appears in a long event log and on which line. This exercises finding a specific entry in a large file and reporting its precise location.
The user asks to append a supplied release-related task as the final line of an existing task list. This exercises preserving existing content while adding exact text in the requested position.
The user asks for the total time a network function spends sleeping when all attempts fail, worked out from the code. This exercises following retry logic and adding the delays along that failure path.
The user asks for a project readme listing every command-line option with its help text. This exercises discovering the program's available options and transferring that information into complete, source-grounded documentation.
The user asks to undo the latest Git commit's change by editing the file directly, without using Git to make changes. This exercises understanding a previous edit and reversing it under a method constraint.
basic-H8 · primary
Filters change this list, not the fixed scoring denominator. Search currently covers question IDs, titles and available task explanations. Focus a control to load prompt/outcome search where this edition provides it.
Suggested comparisons
These are starting points for inspection, not claims that one setup is better. Runtime and placement differences remain visible.
Thinking on versus off
Verified requested on/off conditions of the same configuration, matched on the recorded test protocol.
Requested thinking on · Complete compared coverage