# Model and Effort Guidance
Updated: 2026-10-15. Source: https://robbagley-afk.github.io/model-value-map/
Revision: 5e991ea972cbba3ff5772dd559d754964812d2abf14d41ad733eaba1e63ae58a

## How to use
Use the section for the platform you run on. For each kind of task, recommend the listed pick first.
Move to the listed upgrade only when the pick fails a check or the task is harder than the examples.
Scores and costs come from the sources at the end of this file. Refresh installed instructions by
comparing Revision, not the Updated date. Revision hashes this guidance content, so same-day changes
are detected. These public guides are read-only sources for consumers. Changes reach this website
only after its maintainer builds, validates, commits and pushes the source repository. Editing a
separate note does not publish here.

## Rule
An upgrade pays when it adds at least 5 points (or 75 Elo on GDPval) for each doubling of cost per task. Below that, the cheaper setting is the sweet spot.

## Overall general purpose recommendation
GPT-6.1 Sol at medium effort. Reviewed 2026-10-15. Balanced capability and cost for everyday, multistep work. Use the task-specific picks below when your work needs something different. This is the maintained general-purpose pick, not a universal benchmark winner. A review is due after seven days. Source: https://artificialanalysis.ai/models/gpt-6-1-sol-medium

## Claude (Anthropic) - for Claude Code and Claude apps
- **Mechanical text work** (Classify, extract, reformat, dedupe, parse): Haiku 5.5 low ($0.02, 29). Upgrade: Haiku 5.5 medium ($0.05, 34) if low misses.
- **Scripted relays and tool workflows** (Fixed-step handoffs, connector runs, queue imports): Sonnet 5.5 medium ($0.48, 55%). Opus 5.5 low ($0.55, 53%) when the run touches names or facts. Upgrade: Opus 5.5 medium ($1.34, 61%) when the workflow branches.
- **Documents from supplied facts** (Drafts, summaries, slides, spreadsheets, reports): Haiku 5.5 max (pilot): 1618 at $0.21, near Sonnet 5.5 xhigh at a tenth of the cost. Upgrade: Rarely. Sonnet 5.5 xhigh (1728, $2.01) adds 110 Elo for 9.6x the cost.
- **People, messages, money, records** (Customer and community messages, payments, exact names): Opus 5.5 low ($0.55, 39). Upgrade: Rarely. Opus 5.5 max adds 7 points for 11x the cost.
- **Computer use and browser work** (Clicking through web apps, forms, portals): Haiku 5.5 xhigh (pilot, 67.5%, $0.30). Upgrade: Haiku 5.5 max (72.4%, $0.62) ties Sonnet 5.5 high (73%, $1.50). Sonnet 5.5 xhigh (81%, $2.30) for must-succeed runs.
- **Agentic coding** (Multi-file changes, debugging, repo work): Opus 5.5 medium ($1.34, 53%). Upgrade: Opus 5.5 high ($1.82, 57%) for hard bugs.
- **Small, clear coding fixes** (Clear-repro bug fixes, UI polish, building to a spec): Haiku 5.5 max (pilot, $0.21, 33%), ahead of Opus 5.5 low (31%, $0.55). Upgrade: Opus 5.5 medium if the fix stalls.
- **Research and reasoning** (Analysis, tradeoffs, hard questions): Opus 5.5 low ($0.55, 48%). Vendor chart shows Sonnet 5.5 high at 48% for $0.08 per question. Upgrade: Opus 5.5 high ($1.82, 56%).
- **General agentic judgment** (Open-ended multistep tasks, planning): Opus 5.5 medium ($1.34, 51). Upgrade: Opus 5.5 high ($1.82, 54).

## Codex (OpenAI) - for Codex and ChatGPT
- **Mechanical text work** (Classify, extract, reformat, dedupe, parse): GPT-6 Luna max ($0.07, 38), best when the step calls tools (AutomationBench 53%). Upgrade: GPT-6.1 Sol low ($0.13, 42).
- **Scripted relays and tool workflows** (Fixed-step handoffs, connector runs, queue imports): GPT-6.1 Sol medium ($0.21, 63%). Upgrade: Rarely. GPT-6.1 Sol xhigh ($0.39, 67%) adds 4 points.
- **Documents from supplied facts** (Drafts, summaries, slides, spreadsheets, reports): GPT-6 Luna max (1432, $0.07). Upgrade: GPT-6.1 Sol max (1594, $0.72).
- **People, messages, money, records** (Customer and community messages, payments, exact names): GPT-6.1 Sol high ($0.32, 41), verify names and figures. Upgrade: Rarely. Sol max ($0.72, 42) adds 1 point.
- **Computer use and browser work** (Clicking through web apps, forms, portals): GPT-6 Luna max (49%, $0.22). Sol and Astra not yet evaluated on this chart.
- **Agentic coding** (Multi-file changes, debugging, repo work): GPT-6.1 Sol xhigh ($0.39, 54%). Upgrade: Rarely. GPT-6.1 Sol max ($0.72, 56%) adds 2 points.
- **Small, clear coding fixes** (Clear-repro bug fixes, UI polish, building to a spec): GPT-6.1 Sol medium ($0.21, 48%). Upgrade: GPT-6.1 Sol xhigh ($0.39, 54%).
- **Research and reasoning** (Analysis, tradeoffs, hard questions): GPT-6.1 Sol medium ($0.21, 50%). Upgrade: GPT-6 Astra max ($3.26, 55%), rarely worth it.
- **General agentic judgment** (Open-ended multistep tasks, planning): GPT-6.1 Sol medium ($0.21, 48). Upgrade: GPT-6.1 Sol high ($0.32, 50).

## Gemini (Google)
- **Mechanical text work** (Classify, extract, reformat, dedupe, parse): Gemini 3.8 Flash low (index 33, cost per task not published), also for deterministic scripts. Upgrade: Medium ($0.93, 40).
- **Scripted relays and tool workflows** (Fixed-step handoffs, connector runs, queue imports): Gemini 3.8 Flash medium ($0.93, 61%). Upgrade: None. High scores 60%.
- **Documents from supplied facts** (Drafts, summaries, slides, spreadsheets, reports): Gemini 3.8 Flash medium (1421, $0.93). Upgrade: High adds 15 Elo, not worth it.
- **People, messages, money, records** (Customer and community messages, payments, exact names): Gemini 3.8 Flash medium ($0.93, 29), below Opus 5.5 low and Sol low.
- **Computer use and browser work** (Clicking through web apps, forms, portals): Not yet evaluated.
- **Agentic coding** (Multi-file changes, debugging, repo work): Gemini 3.8 Flash medium ($0.93, 20%) for agent loops, refactors, and hard bugs. Weak for coding. Upgrade: High also scores 20%.
- **Small, clear coding fixes** (Clear-repro bug fixes, UI polish, building to a spec): Gemini 3.8 Flash medium ($0.93, 20%).
- **Research and reasoning** (Analysis, tradeoffs, hard questions): Gemini 3.8 Flash high ($1.24, 48%). Upgrade: None. High is the pick here (HLE 48% vs 42% at medium).
- **General agentic judgment** (Open-ended multistep tasks, planning): Gemini 3.8 Flash medium ($0.93, 40). Upgrade: High adds 1 point, not worth it.

## Grok (xAI)
- **Mechanical text work** (Classify, extract, reformat, dedupe, parse): Grok 4.6 low ($0.38, 35). Upgrade: Grok 4.6 medium ($1.14, 43).
- **Scripted relays and tool workflows** (Fixed-step handoffs, connector runs, queue imports): Grok 4.6 high ($1.48, 67%).
- **Documents from supplied facts** (Drafts, summaries, slides, spreadsheets, reports): Grok 4.6 medium (1612, $1.14). Upgrade: Grok 4.7 high (1706, $2.73).
- **People, messages, money, records** (Customer and community messages, payments, exact names): Grok 4.6 low ($0.38, 26), verify names and figures. Upgrade: Rarely. Grok 4.7 xhigh ($3.74, 32) adds 6 points.
- **Computer use and browser work** (Clicking through web apps, forms, portals): Not yet evaluated.
- **Agentic coding** (Multi-file changes, debugging, repo work): Grok 4.6 high ($1.48, 21%). Upgrade: Rarely. Grok 4.7 high ($2.73, 25%) adds 4 points.
- **Small, clear coding fixes** (Clear-repro bug fixes, UI polish, building to a spec): Grok 4.6 high ($1.48, 21%). Upgrade: Rarely. Grok 4.7 high ($2.73, 25%) adds 4 points.
- **Research and reasoning** (Analysis, tradeoffs, hard questions): Grok 4.6 medium ($1.14, 42%). Upgrade: Rarely. Grok 4.6 xhigh ($1.75, 44%) adds 2 points.
- **General agentic judgment** (Open-ended multistep tasks, planning): Grok 4.6 medium ($1.14, 43). Upgrade: Rarely. Grok 4.7 high ($2.73, 46) adds 3 points.

## Local model (LM Studio)
- **Mechanical text work** (Classify, extract, reformat, dedupe, parse): Gemma 4 26B-A4B ($0). In-house test: classification 6/6, reformatting 8/8, instruction following 6/6 on all three runs. Upgrade: Use Haiku 5.5 low when output must be valid JSON. Gemma broke JSON on 1 to 2 of 6 extractions after audit.
- **Scripted relays and tool workflows** (Fixed-step handoffs, connector runs, queue imports): Light use only. Tool calls 3/4 in-house, AutomationBench 2% (AA estimate). Validate every call. Upgrade: Sonnet 5.5 low.
- **Documents from supplied facts** (Drafts, summaries, slides, spreadsheets, reports): Reformatting only. GDPval 566 (AA estimate).
- **People, messages, money, records** (Customer and community messages, payments, exact names): Never for recall. AA-Omniscience -51. It did answer UNKNOWN to 8/8 made-up facts when offered that option.
- **Computer use and browser work** (Clicking through web apps, forms, portals): Not yet evaluated.
- **Agentic coding** (Multi-file changes, debugging, repo work): Gemma: Terminal-Bench 2.0, our run, 4-bit, thinking off: 15% (13 of 89). Older test, not on this chart.
- **Small, clear coding fixes** (Clear-repro bug fixes, UI polish, building to a spec): Not yet evaluated.
- **Research and reasoning** (Analysis, tradeoffs, hard questions): Not suitable. HLE 8.7% (Google card), money math 4/6 in-house.
- **General agentic judgment** (Open-ended multistep tasks, planning): Not suitable (index 17).

## Best value across all vendors
- **Mechanical text work**: Gemma on a local LM Studio server ($0 cloud). In the cloud, Haiku 5.5 low ($0.02, index 29). Not worth it: Sonnet or Opus for pure mechanical work. Gemma for anything that calls tools (AutomationBench 2%). Grok for mechanical work: Grok 4.6 low ($0.38, 35) costs 19x Haiku 5.5 low for 6 more points. Gemini 3.8 Flash medium ($0.93, 40) costs 13x GPT-6 Luna max for 2 more points. With Gemini, use Flash low (33) for deterministic scripts Newly measured Luna low ($0.0045, index 22) is a candidate for narrow supplied-input steps after a task check, not an established replacement for the maintained picks..
- **Scripted relays and tool workflows**: GPT-6 Luna max ($0.07, AutomationBench 53%), which matches Opus 5.5 low at 13% of the cost. Argon high (78%, $1.99 introductory through November 30) is the highest tracked AutomationBench score under $3; eligibility and task fit still need checking. Not worth it: Haiku 5.5 at any effort (36% at xhigh). Opus 5.5 high and up, Sonnet 5.5 xhigh and max: 1 to 2 points per cost doubling. Gemini 3.8 Flash high (60%) scores below medium ($0.93, 61%), which already trails GPT-6.1 Sol medium (63%, $0.21) Argon high costs $1.99 per AA task through November 30; estimated $3.98 from December 1 assuming unchanged tokens. Subscription caps and availability are separate..
- **Documents from supplied facts**: Haiku 5.5 max (GDPval 1618, $0.21). GPT-6 Luna max (1432, $0.07) if cost matters more than polish. Not worth it: Opus 5.5 high (1704) adds 86 Elo for 8.7x the cost. Sonnet 5.5 max. Haiku hallucinates often (AA-Omniscience 6 to 11), so supply the facts and check names and numbers. Grok 4.6 medium (1612, $1.14) roughly ties Haiku 5.5 max (1618, $0.21) at 5x the cost. Gemini 3.8 Flash medium (1421, $0.93) trails GPT-6 Luna max (1432, $0.07).
- **People, messages, money, records**: GPT-6.1 Sol high ($0.32, AA-Omniscience 41). Sol low ($0.13, 38) for a cheaper first pass with verification. Not worth it: Haiku 5.5 (3 to 11) and Sonnet 5.5 (19 to 32) at any effort. Grok (26 to 32) and Gemini 3.8 Flash (29 to 30) at any effort trail GPT-6.1 Sol low (38, $0.13) Argon high costs $1.99 per AA task through November 30; estimated $3.98 from December 1 assuming unchanged tokens. Subscription caps and availability are separate..
- **Computer use and browser work**: Haiku 5.5 xhigh (OSWorld 67.5% at $0.30 per attempt). Only Anthropic and Luna are on this chart. Not worth it: Sonnet 5.5 max (+2.9 points for 2.5x). Haiku for steps that type names, accounts, or payments.
- **Agentic coding**: GPT-6.1 Sol xhigh (Terminal-Bench 54%, $0.39), ahead of Opus 5.5 medium (53%) at 29% of its cost. Not worth it: Opus 5.5 xhigh (+3 for 1.9x) and max (+0). Sonnet 5.5 max ($5.46). Grok at any effort (3 to 26%): Grok 4.7 xhigh ($3.74, 26%) trails GPT-6.1 Sol low ($0.13, 31%). Gemini 3.8 Flash scores 20% at medium and high. With Gemini, use Flash medium for loops, refactors, and hard bugs, since high adds 33% to the cost for no gain. Argon high costs $1.99 per AA task through November 30; estimated $3.98 from December 1 assuming unchanged tokens. Subscription caps and availability are separate..
- **Small, clear coding fixes**: GPT-6.1 Sol low ($0.13, Terminal-Bench 31%). Haiku 5.5 max ($0.21, 33%) in Claude. Not worth it: Opus 5.5 low for this work, since Haiku 5.5 max scores higher for 38% of the cost. Grok 4.7 low ($1.25, 16%) and Gemini 3.8 Flash medium ($0.93, 20%) trail Haiku 5.5 xhigh ($0.12, 29%).
- **Research and reasoning**: GPT-6.1 Sol medium (HLE 50%, $0.21). Flash high (48%, $1.24) beats Flash medium (42%). Argon high is newly measured at 57% for $1.99 introductory; verify access before adopting it. Not worth it: Opus 5.5 xhigh and max: 2 to 3 points per cost doubling. Grok tops out at 44% (Grok 4.6 xhigh, $1.75), below GPT-6.1 Sol low (47%, $0.13) Argon high costs $1.99 per AA task through November 30; estimated $3.98 from December 1 assuming unchanged tokens. Subscription caps and availability are separate..
- **General agentic judgment**: GPT-6.1 Sol medium (index 48, $0.21). Not worth it: Opus 5.5 xhigh and max (+2 per doubling). Sonnet 5.5 high (47, $0.88) trails Opus 5.5 medium on every test. Grok 4.7 high and xhigh (46) cost 13 to 18x GPT-6.1 Sol medium for a lower score. Gemini 3.8 Flash high adds 1 point over medium (41 vs 40) for 33% more. Fable 5.1 high (51, $3.91) ties Opus 5.5 medium (51, $1.34) at nearly 3x the task cost. Argon high costs $1.99 per AA task through November 30; estimated $3.98 from December 1 assuming unchanged tokens. Subscription caps and availability are separate..

## Latest news
Dated news is a snapshot. Ignore an item after its expiry; offers do not establish subscription savings.
- 2026-10-15 **Argon introductory API pricing**: $2 input / $10 output per million tokens through November 30, then $4 / $20. API offer only; subscription allowances are separate. Expires 2026-12-01T00:00:00Z. Checked 2026-10-15. Source: https://ai.google.dev/gemini-api/docs/pricing?hl=en
- 2026-10-13 **Gemini 4 Argon broadly available**: Independent AutomationBench: 78% versus Sol Medium 63%, at $1.99 versus $0.21 per benchmark task. More capability comes at a higher task cost. Expires 2026-11-12T00:00:00Z. Checked 2026-10-15. Source: https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon/ Independent evidence: https://artificialanalysis.ai/models/comparisons/gemini-4-argon-vs-gpt-6-1-sol-medium
- 2026-10-07 **Sonnet 5.5 cache reads halved**: Cache reads fell to $0.10 per million tokens. Anthropic estimates about 20% lower cost on most agentic tasks; this is a vendor claim, not a measured subscription saving. Expires 2026-11-06T00:00:00Z. Checked 2026-10-15. Source: https://www.anthropic.com/claude-haiku-5-5

## Promotional pricing and expiry
Prices are date-dependent. Read the promotion end time before using a copied price. The live maps recheck the clock on load, once daily, and when a tab becomes visible. Static guidance is a dated snapshot; after a boundary, use the standard estimates below until the maintainer reviews recommendations. No subscription allowance or account-specific credits are assumed.
- **Gemini 3.8 Flash**: 50% introductory API pricing through 2026-12-31; expires 2027-01-01T00:00:00Z. Scope: Google AI Studio / Gemini Developer API paid Standard tier; not a subscription-plan discount. Checked 2026-10-15.
  USD per million input/output/cached-input tokens: 0.75 / 3.75 / 0.075; standard from 2027-01-01: 1.5 / 7.5 / 0.15.
  Conservative UTC cutoff: vendor lists dates but does not specify a timezone. Use January 1 at 00:00 UTC as a conservative display cutoff; billing follows vendor terms. AA reported medium/high costs already use introductory input/output/cache rates. Standard task estimates assume identical token usage; not a fresh benchmark run. Source: https://ai.google.dev/gemini-api/docs/pricing
  medium: reported introductory cost $0.93/task; estimated standard cost $1.86/task. Do not halve the reported cost again.
  high: reported introductory cost $1.24/task; estimated standard cost $2.48/task. Do not halve the reported cost again.
- **Gemini 4 Argon**: 50% introductory API pricing through 2026-11-30; expires 2026-12-01T00:00:00Z. Scope: Gemini Developer API paid Standard tier. No free tier. Subscription eligibility and Antigravity allowances are separate.. Checked 2026-10-15.
  USD per million input/output/cached-input tokens: 2 / 10 / 0.1; standard from 2026-12-01: 4 / 20 / 0.2.
  Conservative UTC display cutoff because the vendor states December 1 without a timezone; vendor billing terms remain authoritative. Exact offer start timestamp is not specified. AA high-effort cost $1.99 already includes introductory rates; post-expiry $3.98 assumes unchanged token usage and is an estimate, not a fresh benchmark measurement. Source: https://ai.google.dev/gemini-api/docs/pricing?hl=en
  high: reported introductory cost $1.99/task; estimated standard cost $3.98/task. Do not halve the reported cost again.

## Subscription usage offers (not API discounts)*
Goal: maximize useful completed work within the applicable usage limit. API cost per benchmark task is not a measurement of subscription allowance. A 50% allowance increase corresponds to about 33% lower effective cost per equal unit; 50% less counted usage can double eligible capacity. Weekly caps, eligibility and task quality still constrain value.
- **Claude artifact usage promotion***: 50% less five-hour session usage for eligible artifact work; approximately 2x eligible session capacity, not a blanket token discount. Ends October 15, 2026 at 11:59 PM Pacific (2026-10-16T07:00:00Z). Checked 2026-10-15.
Free (from October 7), Pro, Max and Team; excludes Enterprise. Eligible Claude chats and cloud tasks in the Claude app only.
Chat: next 10 messages after an artifact edit, up to the first 15 steps per reply. Cloud tasks in the Claude app: about 45 minutes when starting with Document, Presentation or Design, or up to 80 steps after an artifact edit.
For eligible document, slide and design work, consider Claude during this offer to stretch the five-hour allowance. Choose the lowest model and effort that passes your checks. API benchmark costs do not measure subscription allowance; no cross-vendor subscription winner is established.
* Does not apply to API use. No Claude API, Claude Code, local desktop tasks in the Claude app, scheduled/unattended tasks, usage-credit charges, older artifacts, standalone Claude Design, or Google Drive-created Docs/Sheets/Slides. Weekly allowance is unchanged. Source: https://support.claude.com/en/articles/17274727-artifact-usage-promotion
After expiry, standard usage applies; ignore this temporary recommendation.

## Sources
- https://artificialanalysis.ai/models/comparisons/claude-fable-5-1-low-vs-gpt-6-1-sol-medium
- https://artificialanalysis.ai/models/comparisons/claude-fable-5-1-medium-vs-gpt-6-1-sol-medium
- https://artificialanalysis.ai/models/comparisons/claude-fable-5-1-high-vs-gpt-6-1-sol-medium
- https://artificialanalysis.ai/models/comparisons/claude-fable-5-1-xhigh-vs-gpt-6-1-sol-medium
- https://artificialanalysis.ai/models/comparisons/claude-fable-5-1-vs-gpt-6-1-sol-medium
- https://artificialanalysis.ai/models/comparisons/claude-haiku-5-5-low-vs-claude-4-5-haiku
- https://artificialanalysis.ai/models/comparisons/claude-haiku-5-5-medium-vs-gpt-6-1-sol-medium
- https://artificialanalysis.ai/models/comparisons/claude-haiku-5-5-high-vs-claude-sonnet-5-5-xhigh
- https://artificialanalysis.ai/models/comparisons/claude-haiku-5-5-xhigh-vs-claude-sonnet-5-5-low
- https://artificialanalysis.ai/models/comparisons/claude-haiku-5-5-vs-claude-opus-5-5-low
- https://artificialanalysis.ai/models/comparisons/claude-sonnet-5-5-medium-vs-claude-opus-5-5-low
- https://artificialanalysis.ai/models/comparisons/claude-sonnet-5-5-high-vs-claude-opus-5-5-medium
- https://artificialanalysis.ai/models/comparisons/claude-opus-5-5-high-vs-claude-opus-5-5-xhigh
- https://artificialanalysis.ai/models/comparisons/claude-sonnet-5-5-vs-claude-opus-5-5
- https://artificialanalysis.ai/models/comparisons/gpt-6-astra-low-vs-gpt-6-1-sol-medium
- https://artificialanalysis.ai/models/comparisons/gpt-6-astra-medium-vs-gpt-6-1-sol-medium
- https://artificialanalysis.ai/models/comparisons/gpt-6-astra-high-vs-gpt-6-1-sol-medium
- https://artificialanalysis.ai/models/comparisons/gpt-6-astra-xhigh-vs-gpt-6-1-sol-medium
- https://artificialanalysis.ai/models/comparisons/gpt-6-luna-vs-gpt-6-astra
- https://artificialanalysis.ai/models/comparisons/gpt-6-luna-low-vs-gpt-6-1-sol-medium
- https://artificialanalysis.ai/models/comparisons/gpt-6-luna-medium-vs-gpt-6-1-sol-medium
- https://artificialanalysis.ai/models/comparisons/gpt-6-luna-high-vs-gpt-6-1-sol-medium
- https://artificialanalysis.ai/models/comparisons/gpt-6-luna-xhigh-vs-gpt-6-1-sol-medium
- https://artificialanalysis.ai/models/comparisons/gpt-6-1-sol-low-vs-claude-sonnet-5-5-low
- https://artificialanalysis.ai/models/comparisons/gpt-6-1-sol-high-vs-gpt-6-1-sol
- https://artificialanalysis.ai/models/comparisons/gpt-6-1-sol-xhigh-vs-gpt-6-1-sol-medium
- https://artificialanalysis.ai/models/comparisons/gemini-3-8-flash-low-vs-gemini-3-8-flash-medium
- https://artificialanalysis.ai/models/comparisons/gemini-3-8-flash-vs-gemini-3-8-flash-medium
- https://artificialanalysis.ai/models/comparisons/grok-4-6-low-vs-grok-4-6-xhigh
- https://artificialanalysis.ai/models/comparisons/grok-4-7-low-vs-grok-4-6-medium
- https://artificialanalysis.ai/models/comparisons/grok-4-7-vs-grok-4-6
- https://artificialanalysis.ai/models/comparisons/grok-4-7-high-vs-gpt-6-1-sol-medium
- https://artificialanalysis.ai/models/comparisons/gemma-4-26b-a4b-vs-gemini-3-8-flash-low
- https://artificialanalysis.ai/models/comparisons/gemini-4-argon-vs-gpt-6-1-sol-medium

## Review status
All 47 tracked independent model/effort points verified from comparison tables on 2026-10-15. Added 14 previously missing effort points. GDPval Elo can move as comparison results update. Vendor chart points remain approximate and dated 2026-10-08. Unpublished costs and benchmark scores remain null. Independent reasoning scores do not describe the local non-thinking QAT instance. Overall and task-specific model picks retained after review. Account-specific offers without verified eligibility and complete dates are excluded from price calculations. The newly added model has one verified effort point; other attempted effort sources remain pending. Existing task-model picks remain in place pending the new model access and effort comparison review; news and charts do not assume account eligibility.
Pending source refresh: Luna none: attempted comparison returned HTTP 404, Grok 4.7 medium: attempted comparison returned HTTP 404, Vendor chart images: approximate points retained from 2026-10-08, OpenAI Pro 200 temporary allowance: official terms confirm eligible subscriptions active September 22 through September 29 at 10 AM Pacific keep the previous allowance through October 29. Exact promotion start and expiry timezone, specific account eligibility and any credit grant are unverified. No API discount or credit amount assumed. Source checked 2026-10-15: https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers, OpenAI referral rewards vary by account and offer. Banked Work/Codex resets normally expire 30 days after award; exact account offer, caps and award date are unverified. Not API credits. No price adjustment. Source checked 2026-10-15: https://help.openai.com/en/articles/20001271-chatgpt-desktop-referral-promotions, Google Antigravity promotional overage credits require account-specific terms. No verified temporary subscription boost found in reviewed official plans/pricing. No API discount inferred. Sources checked 2026-10-15: https://antigravity.google/docs/plans and https://antigravity.google/pricing, Grok for Startups may cover API and other eligible services only after approval. Amount and expiry come from approval terms, normally six months after credit application; account eligibility and grant are unverified. Cannot stack without written approval. No token-price adjustment. Source checked 2026-10-15: https://x.ai/legal/startup-program-terms, Gemini 4 Argon low/medium: attempted comparison URLs returned HTTP 404; no scores inferred.