Generalist / code / analysis / agents

OpenAI (GPT)

  • GPT-6 Astra from 09/03/2026 $10/$50

    Computer use, coding and office work. ChatGPT, the API (gpt‑6‑astra) and Bedrock. Off in Enterprise until an admin turns it on.

  • GPT-6.1 Sol from 09/29/2026 $2/$10

    Work, Codex and the API (gpt‑6.1‑sol) on Plus, Pro, Business, Enterprise and Edu. Not in regular chat yet.

  • GPT-5.6 from 07/30/2026 from $0.20

    Luna $0.20/$1.20, Terra $2/$12. Sol promo $4/$20 through 11/21/2026.

Voice: GPT-Live from 07/08/2026. Self-hosting: gpt-oss, Apache 2.0.

  • Verified2026-09-30
  • Models covered7
  • Sources11
  • Deployment channels7

Purchase decision (when to choose / when to avoid)

Choose if...

  • You need computer use, long agentic workflows or frontier-level coding - GPT-6 Astra (from 09/03/2026).
  • You want coding, computer use or professional work close to Astra at $2/$10 - GPT-6.1 Sol in ChatGPT Work, Codex and the API (from 09/29/2026).
  • You want a lower-cost default for daily work - GPT-5.6 Terra/Luna in ChatGPT Work.
  • You need voice conversations in ChatGPT (hands-free, live translation) - GPT-Live from 07/08/2026.
  • You build on API/Codex: Astra for the hardest tasks, 6.1 Sol for coding and computer use, Terra/Luna for volume with predictable cache.
  • You need a mature ecosystem (integrations, tools, stable API, versioning).
  • You're considering open-weight (gpt-oss) as an alternative for self-hosting in selected processes.

Avoid if...

  • You want to avoid vendor lock-in and have full weight control (prefer open-weight).
  • Token cost is critical at high volume - Terra and Luna cost less than Astra ($10/$50) and less than 6.1 Sol ($2/$10).
  • You are moving from gpt-6-sol: 6.1 Sol has no reasoning none or minimal, and API tools require the Responses API.
  • You want 6.1 Sol in regular ChatGPT chat: at launch it is in Work and Codex, not in Chat.
  • You need Astra in Enterprise without an admin decision - off by default at launch.
  • You need voice in ChatGPT Business/Enterprise/Edu - GPT-Live not available at launch.
  • You have very restrictive data/region requirements - compare terms (Azure OpenAI / enterprise, API ZDR).

Cost in practice (scenarios)

SMB / marketing 1-10 people

Usually: ChatGPT Team + optional API for automation.

  • copy, analytics, presentations
  • moderate volume
Company 10-200 people

Enterprise if you need SSO, data policies and control; Astra requires admin enable.

  • multiple departments
  • central IT governance
GPT-6.1 Sol (ChatGPT Work / Codex / API)

$2/$10 per 1M tok. in the API (cache read $0.10, cache write $2.50). Above 272k input tokens: 2x input and cache, 1.5x output.

  • from 09/29/2026, id gpt-6.1-sol
  • Work and Codex: Plus/Pro/Business/Enterprise/Edu; regular chat does not include it yet
  • API tools only through Responses; Fast mode does not work with EU residency
GPT-6 Astra (ChatGPT / API / Codex)

$10/$50 per 1M tok. in the API (cache read $1); on Plus+ plans within subscription allowances, optional extra credits.

  • staged rollout from 09/03/2026
  • Fast mode 2x speed at 2x price
  • Pro/Business/Enterprise: GPT-6 Astra Pro
GPT-5.6 (ChatGPT Work / API / Codex)

Sol $4/$20 (promo through 11/21/2026; list $5/$30), Terra $2/$12, Luna $0.20/$1.20 per 1M tok. (from 07/30/2026); Free/Go: Terra; Plus+: Sol/Terra/Luna choice.

  • GA from 07/09/2026
  • volume and daily work line
ChatGPT Voice (GPT-Live)

Included in Free/Go/Plus/Pro plans; no separate pricing at launch.

  • iOS/Android/web
  • Business/Enterprise/Edu not included at launch
Automations (API)

Cost depends on tokens; cache, batch and context shortening pay off.

  • RAG, webhooks, integrations
  • Astra for the hardest tasks, 6.1 Sol for coding and computer use, Terra/Luna for volume
These are estimates/scenarios (not an invoice). Actual cost depends on context length, number of users, limits and retention policies.

Deployment / data / enterprise

Deployment channels

  • ChatGPT (Free/Go/Plus/Pro/Business/Enterprise) - GPT-6 Astra staged from 09/03/2026
  • ChatGPT Work and Codex - GPT-6.1 Sol from 09/29/2026 (Plus/Pro/Business/Enterprise/Edu); regular chat does not include 6.1 Sol
  • ChatGPT Voice (GPT-Live)
  • OpenAI API (gpt-6-astra, gpt-6.1-sol, GPT-5.6 Sol/Terra/Luna)
  • Amazon Bedrock (GPT-6 Astra)
  • Azure OpenAI Service
  • Self-host: gpt-oss (open-weight)

Data policy

Training on data
API/Enterprise: not by default (check terms in sources).
Retention
Depends on service (API vs ChatGPT vs Azure); GPT-6 Astra: ZDR for eligible API customers.
Data residency
Depends on region and provider (OpenAI / Azure / Bedrock).
In companies, API or ChatGPT Enterprise is usually chosen for data policies.

Enterprise readiness

Admin
ChatGPT Team/Enterprise + admin, policies, workspace; GPT-6 Astra off by default in Enterprise, admin enable required.
Largest ecosystem - fastest time-to-value, but watch costs at volume (Astra $10/$50 vs Terra $2/$12).

Best use cases

  • computer use and multi-step browser/desktop workflows - GPT-6 Astra
  • long coding sessions in Codex with notes across context windows - GPT-6 Astra
  • coding, computer use and professional work cheaper than Astra - GPT-6.1 Sol in ChatGPT Work, Codex and the API ($2/$10, from 09/29/2026)
  • frontier tasks at a higher API rate - GPT-6 Astra; volume and daily work - GPT-5.6 Terra/Luna
  • hands-free voice conversations in ChatGPT (GPT-Live: full duplex, live translation)
  • ecosystem integrations (BI, Notion, Slack, Microsoft, Azure OpenAI) and long-horizon agent ops (MCP)
  • self-host open-weight (gpt-oss 120B/20B) when control is needed and MLOps competence exists.

Strengths

  • GPT-6.1 Sol (09/29/2026): $2/$10, cache read $0.10, 1.05M context. OpenAI describes quality close to Astra. Input and output rates are one fifth of Astra ($2/$10 versus $10/$50).
  • GPT-6 Astra: SOTA computer use (OSWorld 2.0 72.6%), coding (Terminal-Bench 4.0 57.9%), better intent alignment than GPT-5.6 Sol.
  • 1.05M context with strong retrieval (96.3% MRCR 512K-1M); notes across context windows in Codex.
  • Very mature API, deprecation policy and documentation; widest integration ecosystem.
  • GPT-5.6: Sol/Terra/Luna in ChatGPT Work, Codex and API; Terra/Luna for volume after the 07/30/2026 price cut.
  • GPT-Live: natural full-duplex voice; gpt-oss: Apache 2.0, MoE, self-host.

Weaknesses / risks

  • Astra is more expensive in the API ($10/$50) than GPT-5.6 Terra/Luna; staged rollout, off by default in Enterprise.
  • Possible pauses and extra safety checks on cyber tasks (defensive and offensive).
  • No weights for closed models → vendor lock-in; cost at high volume; risk of behavior changes (versioning).
  • GPT-Live not available for Business/Enterprise/Edu workspaces; ZDR in the API only for eligible customers.

Current models (examples)

  • GPT-6 Astra (09/03/2026) - frontier: computer use, coding, cyber, science, professional work; 1.05M context, 128k output; ChatGPT Plus/Pro/Business/Enterprise (staged), API: gpt-6-astra, Amazon Bedrock; Pro/Business/Enterprise: GPT-6 Astra Pro; Enterprise: admin enable required.
  • GPT-6 Astra API: $10/$50 per 1M tok., cache read $1, cache write $12.50; Fast mode 2x speed at 2x price.
  • GPT-6.1 Sol (09/29/2026) - ChatGPT Work and Codex (Plus/Pro/Business/Enterprise/Edu), not in regular chat; API gpt-6.1-sol; $2/$10, cache read $0.10, cache write $2.50; 1.05M context, 128k output; above 272k input: 2x input and cache, 1.5x output; Fast 2x, Batch and Flex at half the standard rate; US and EU residency, Fast unavailable with EU residency.
  • GPT-5.6 Sol / Terra / Luna (GA from 07/09/2026) - ChatGPT Work, Codex, API; Terra default Free/Go; Sol/Terra/Luna on Plus+; from 07/30/2026 Terra $2/$12, Luna $0.20/$1.20; Sol promo $4/$20 through 11/21/2026 (list $5/$30).
  • GPT-Live-1 / GPT-Live-1 mini (07/08/2026) - full-duplex voice in ChatGPT; Go/Plus/Pro vs Free.
  • GPT-5.5 Pro / Instant (Apr-May 2026) - previous generation; still in the API.
  • gpt-oss (120B/20B) - open-weight, Apache 2.0, on-prem/cloud/edge; 'harmony' format.

Alternatives (if this model doesn't fit)

CL
Claude
safety + long context + agentic capabilities
GE
Gemini
GCP/Vertex, multimodality, ZDR in config
DE
DeepSeek
cheap reasoning/code at high volume