Models

The labs keep shipping. This page is the current flagship cut: closed APIs from Google, Anthropic, OpenAI, Meta, Moonshot, xAI, and Alibaba, plus the open-weight models that actually sit next to them on a shared harness — DeepSeek, GLM, MiMo, MiniMax, Mistral, Muse Spark Open, Kimi, Qwen, Gemma. Quality is the Artificial Analysis Intelligence Index v5.0, not a press-release table. v5.0 is a new, harder composite: scores run much lower than the v4.x numbers we used before and are not comparable to them.

Prices are USD per million tokens on the first-party (or AA-measured) API; Gemini rows show Google's introductory list prices. If a field is missing, it is a dash, not a guess. Rows with a dash for Index are current downloadable models AA has not yet scored on v5.0. We do not treat every 7B fine-tune as a frontier model.

Price vs valueAPI tiers: list $/1M next to what you get.Plan tiersChat/product subscriptions: Free → Enterprise.ArenaMultiple Arena boards as charts — Text, Code, Vision, Search, Agent.LiveBenchCategory charts + full table — contamination-resistant.Pricing historyInput and output $/1M over time — step chart.Usage AdvisorReprice your API mix; CSV stays in-browser.What ifTools: past stock buys.

Artificial Analysis Intelligence Index v5.0

Higher is better. Navy is frontier / API-only. Accent is open weight. Effort settings are the ones AA scored, not always the lab default. Bars are scaled to this board's score range (not from zero).

Frontier closedOpen weight

Gemini 4 Carbon59
Claude Opus 5.558
GPT-6 Astra57
GPT-6.1 Sol55
Claude Sonnet 5.554
GPT-6.1 Terra51
Gemini 4 Argon50
Muse Wonder Full50
Claude Fable 5.149
Claude Mythos 5.149
Kimi K3.1 Pro46
Muse Wonder43
Muse Spark 1.336
Claude Haiku 5.536
Grok 4.735
Qwen3.8 Max (0902)34
DeepSeek V4.1 Flash 101732
GLM-5.332
Gemini 3.8 Flash31
MiMo-V2.6-Pro30
MiniMax-M3.1-Flash29
GPT-6 Luna28
Mistral Large 426
Muse Spark Open25
DeepSeek V4 Pro 081323

Filter

Sort

Gemini 4 Carbon
Google · Frontier closed
Google's newest frontier model, built for cybersecurity and long-running agent work; #1 on AA Index v5.0.
Index
59
Context
1M
In / 1M
$2
Out / 1M
$10
License
—
Released
28 Oct 2026
Scored: high
Limited access: cyber defenders via Google's Fairwind Program first; broader developer access not yet dated. Introductory list; $4 / $20 after the introductory period. Cached input 95% off.
Claude Opus 5.5
Anthropic · Frontier closed
Anthropic's current flagship: roughly Fable 5.1-level work at 20% below Opus 5's list price.
Index
58
Context
1M
In / 1M
$4
Out / 1M
$20
License
—
Released
22 Sep 2026
Scored: max effort, default fallback
Cache reads $0.20 per 1M. Fast mode $8 / $40.
GPT-6 Astra
OpenAI · Frontier closed
OpenAI's GPT-6 flagship for the hardest end-to-end agent work.
Index
57
Context
1.05M
In / 1M
$10
Out / 1M
$50
License
—
Released
3 Sep 2026
Scored: max effort
Short-context list. Long context bills $20 / $75 per 1M. Cached input $1.00.
GPT-6.1 Sol
OpenAI · Frontier closed
Near-Astra quality at a fifth of Astra's list price; replaces GPT-5.6 Sol and GPT-6 Sol.
Index
55
Context
1.05M
In / 1M
$2
Out / 1M
$10
License
—
Released
29 Sep 2026
Scored: max effort
Cached input $0.10 per 1M. Prompts over 272k input tokens bill 2× input.
Claude Sonnet 5.5
Anthropic · Frontier closed
Workhorse Claude: four points off Opus 5.5 on the Index at half the price.
Index
54
Context
1M
In / 1M
$2
Out / 1M
$10
License
—
Released
28 Sep 2026
Scored: max effort, default fallback
Cache reads $0.10 per 1M.
GPT-6.1 Terra
OpenAI · Frontier closed
OpenAI's mid tier, tuned for computer use and visual reasoning; half the price of GPT-6.1 Sol.
Index
51
Context
1.05M
In / 1M
$1
Out / 1M
$5
License
—
Released
26 Oct 2026
Scored: max effort
Cached input $0.05 per 1M.
Gemini 4 Argon
Google · Frontier closed
Google's generally available Gemini 4 frontier model; replaces Gemini 3.1 Pro Preview as the Pro-class pick.
Index
50
Context
1M
In / 1M
$2
Out / 1M
$10
License
—
Released
13 Oct 2026
Scored: high
Introductory list through 30 Nov 2026; then $4 / $20 per 1M. Cached input 95% off. 1M unified input-output limit.
Muse Wonder Full
Meta · Frontier closed
Meta's top model, but only inside a Muse Code subscription.
Index
50
Context
1M
In / 1M
—
Out / 1M
—
License
—
Released
12 Oct 2026
Scored: max
No API token price: Muse Code subscription only (Power Usage and Maximum Usage plans).
Claude Fable 5.1
Anthropic · Frontier closed
Anthropic's highest-priced Claude; on Index v5.0 it now trails the cheaper Opus 5.5.
Index
49
Context
1M
In / 1M
$10
Out / 1M
$50
License
—
Released
1 Sep 2026
Scored: max effort, default fallback
Cache reads $0.25 per 1M tokens (0.025× input).
Claude Mythos 5.1
Anthropic · Frontier closed
Limited-access twin of Fable 5.1 (same weights) with a more permissive cyber/life-sciences safeguard layer.
Index
49
Context
1M
In / 1M
$10
Out / 1M
$50
License
—
Released
1 Sep 2026
Scored: AA scored Fable 5.1 (identical weights); max effort, default fallback
Limited availability — verified organizations / trusted access only. Cache reads $0.25 per 1M tokens.
Kimi K3.1 Pro
Moonshot / Kimi · Frontier closed
Moonshot's post-trained K3; best Chinese model on Index v5.0, ahead of every open-weight row.
Index
46
Context
1M
In / 1M
$2
Out / 1M
$10
License
—
Released
23 Oct 2026
Kimi API price as AA lists it. API-only today; Moonshot says full weights ship by 4 Nov 2026 — move to open weight when they land.
Muse Wonder
Meta · Frontier closed
Meta's new API flagship for agents and computer use; replaces Muse Spark as Meta's top paid model.
Index
43
Context
1M
In / 1M
$4
Out / 1M
$20
License
—
Released
12 Oct 2026
Scored: max
muse-wonder-1.0, Standard mode. Cached input $0.20. No Contributor-mode discount.
Muse Spark 1.3
Meta · Frontier closed
Meta's cheaper API line, now below Muse Wonder.
Index
36
Context
1M
In / 1M
$1.25
Out / 1M
$4.25
License
—
Released
2 Sep 2026
Scored: max
Standard mode (not used for training). Cached input $0.15 per 1M. Contributor mode is $0.10 / $0.20.
Claude Haiku 5.5
Anthropic · Frontier closed
Anthropic's small model for high-volume, latency-sensitive work; first Haiku with an effort setting.
Index
36
Context
1M
In / 1M
$0.10
Out / 1M
$0.50
License
—
Released
7 Oct 2026
Scored: max effort
List for prompts up to 100k tokens; prompts over 100k bill $0.50 / $2.50 per 1M. Cache reads $0.01 / $0.05 per 1M. 128K max output.
Grok 4.7
xAI · Frontier closed
xAI's (now SpaceXAI) current coding and agent flagship.
Index
35
Context
500k
In / 1M
$2
Out / 1M
$6
License
—
Released
21 Sep 2026
Scored: high (default); xhigh scores 34
Under 200k prompt tokens ($0.50 cached); $4 / $12 above. Grok 5 was previewed 14 Oct 2026 but is not in the API.
Qwen3.8 Max (0902)
Alibaba / Qwen · Frontier closed
Alibaba's proprietary API flagship; the highest-scoring Qwen on Index v5.0.
Index
34
Context
984k
In / 1M
$2
Out / 1M
$6
License
—
Released
2 Sep 2026
Alibaba Cloud Model Studio international list.
DeepSeek V4.1 Flash 1017
DeepSeek · Open weight
Top open-weight model on Index v5.0 (tied with GLM-5.3), at a fraction of the price.
Index
32
Context
1M
In / 1M
$0.30
Out / 1M
$1.20
License
MIT
Released
17 Oct 2026
Scored: max effort
First-party API as measured by Artificial Analysis. The deepseek-flash alias routes here.
GLM-5.3
Zhipu / Z.AI · Open weight
Zhipu's open-weight frontier; ties DeepSeek for the open-weight lead on v5.0.
Index
32
Context
1M
In / 1M
$1.40
Out / 1M
$4.40
License
GLM-5 license
Released
18 Aug 2026
Scored: max effort
Gemini 3.8 Flash
Google · Frontier closed
Google's current Flash; still the cheap, fast Gemini until a Gemini 4 Flash ships.
Index
31
Context
1.05M
In / 1M
$0.75
Out / 1M
$3.75
License
—
Released
2 Sep 2026
Scored: high
Introductory Google list through 31 Dec 2026; then $1.50 / $7.50 from 1 Jan 2027.
MiMo-V2.6-Pro
Xiaomi · Open weight
Xiaomi's 1.02T MoE (42B active) with native text, image, video, and audio input; third-best open weights on v5.0.
Index
30
Context
1M
In / 1M
$0.43
Out / 1M
$0.87
License
MIT
Released
21 Sep 2026
API price as AA lists it.
MiniMax-M3.1-Flash
MiniMax · Open weight
428B MoE (23B active) for coding and agents; near the open-weight top for pennies.
Index
29
Context
1M
In / 1M
$0.15
Out / 1M
$0.60
License
MiniMax Community License
Released
19 Oct 2026
Cached input $0.015 per 1M.
GPT-6 Luna
OpenAI · Frontier closed
OpenAI's volume tier; replaces GPT-5.6 Luna at half the price.
Index
28
Context
1.05M
In / 1M
$0.10
Out / 1M
$0.50
License
—
Released
22 Sep 2026
Scored: max effort
Cached input $0.01 per 1M. Prompts over 272k input tokens bill 2× input.
Mistral Large 4
Mistral · Open weight
"Le Chonk": Mistral's 1.05T MoE (52B active) flagship, now open weight; replaces Medium 3.5 here.
Index
26
Context
1M
In / 1M
$0.70
Out / 1M
$2.10
License
Mistral Ecosystem Open-Weight License
Released
29 Oct 2026
Scored: high
Cached input $0.07 per 1M. API public preview from 6 Oct 2026; weights released 29 Oct 2026. Commercial model-as-a-service use needs a separate Mistral license.
Muse Spark Open
Meta · Open weight
Meta's downloadable model now (816B MoE, 42B active); replaces Llama 4 as the Meta open-weight row.
Index
25
Context
1M
In / 1M
$1.25
Out / 1M
$4.25
License
Muse Model License
Released
15 Oct 2026
Scored: xhigh
Hosted-provider price as AA lists it; not on Meta Model API. Weights on Hugging Face.
DeepSeek V4 Pro 0813
DeepSeek · Open weight
1.6T MoE, 49B active, MIT weights; DeepSeek's big model, now behind its own Flash on v5.0.
Index
23
Context
1M
In / 1M
$1.32
Out / 1M
$3.96
License
MIT
Released
13 Aug 2026
Scored: max effort
First-party API as measured by Artificial Analysis. The newer V4.1 Flash 1017 now outscores it.
Kimi K3
Moonshot / Kimi · Open weight
Moonshot's current downloadable model (2.8T MoE, 104B active).
Index
—
Context
1.05M
In / 1M
$3
Out / 1M
$15
License
Kimi K3 License (commercial with restrictions)
Released
16 Jul 2026
Not yet scored on AA Index v5.0. Kept until Kimi K3.1 Pro weights ship (promised by 4 Nov 2026).
Qwen3.8 2.4T A95B
Alibaba / Qwen · Open weight
Alibaba's largest open-weight Qwen (2.4T MoE, 95B active).
Index
—
Context
262k
In / 1M
$2
Out / 1M
$6
License
Qwen3 license
Released
12 Aug 2026
Not yet scored on AA Index v5.0. AA now lists a 262k context for this model.
Gemma 4 31B
Google · Open weight
Google's current Gemma dense open-weight; Apache 2.0, 256k context.
Index
—
Context
256k
In / 1M
—
Out / 1M
—
License
Apache 2.0
Released
2 Apr 2026
Not yet scored on AA Index v5.0. Self-host; no Google list price.

	Lab	Kind					License
Gemini 4 Carbon Google's newest frontier model, built for cybersecurity and long-running agent work; #1 on AA Index v5.0. Scored: high Limited access: cyber defenders via Google's Fairwind Program first; broader developer access not yet dated. Introductory list; $4 / $20 after the introductory period. Cached input 95% off.	Google	Frontier closed	59	1M	$2	$10	—	28 Oct 2026
Claude Opus 5.5 Anthropic's current flagship: roughly Fable 5.1-level work at 20% below Opus 5's list price. Scored: max effort, default fallback Cache reads $0.20 per 1M. Fast mode $8 / $40.	Anthropic	Frontier closed	58	1M	$4	$20	—	22 Sep 2026
GPT-6 Astra OpenAI's GPT-6 flagship for the hardest end-to-end agent work. Scored: max effort Short-context list. Long context bills $20 / $75 per 1M. Cached input $1.00.	OpenAI	Frontier closed	57	1.05M	$10	$50	—	3 Sep 2026
GPT-6.1 Sol Near-Astra quality at a fifth of Astra's list price; replaces GPT-5.6 Sol and GPT-6 Sol. Scored: max effort Cached input $0.10 per 1M. Prompts over 272k input tokens bill 2× input.	OpenAI	Frontier closed	55	1.05M	$2	$10	—	29 Sep 2026
Claude Sonnet 5.5 Workhorse Claude: four points off Opus 5.5 on the Index at half the price. Scored: max effort, default fallback Cache reads $0.10 per 1M.	Anthropic	Frontier closed	54	1M	$2	$10	—	28 Sep 2026
GPT-6.1 Terra OpenAI's mid tier, tuned for computer use and visual reasoning; half the price of GPT-6.1 Sol. Scored: max effort Cached input $0.05 per 1M.	OpenAI	Frontier closed	51	1.05M	$1	$5	—	26 Oct 2026
Gemini 4 Argon Google's generally available Gemini 4 frontier model; replaces Gemini 3.1 Pro Preview as the Pro-class pick. Scored: high Introductory list through 30 Nov 2026; then $4 / $20 per 1M. Cached input 95% off. 1M unified input-output limit.	Google	Frontier closed	50	1M	$2	$10	—	13 Oct 2026
Muse Wonder Full Meta's top model, but only inside a Muse Code subscription. Scored: max No API token price: Muse Code subscription only (Power Usage and Maximum Usage plans).	Meta	Frontier closed	50	1M	—	—	—	12 Oct 2026
Claude Fable 5.1 Anthropic's highest-priced Claude; on Index v5.0 it now trails the cheaper Opus 5.5. Scored: max effort, default fallback Cache reads $0.25 per 1M tokens (0.025× input).	Anthropic	Frontier closed	49	1M	$10	$50	—	1 Sep 2026
Claude Mythos 5.1 Limited-access twin of Fable 5.1 (same weights) with a more permissive cyber/life-sciences safeguard layer. Scored: AA scored Fable 5.1 (identical weights); max effort, default fallback Limited availability — verified organizations / trusted access only. Cache reads $0.25 per 1M tokens.	Anthropic	Frontier closed	49	1M	$10	$50	—	1 Sep 2026
Kimi K3.1 Pro Moonshot's post-trained K3; best Chinese model on Index v5.0, ahead of every open-weight row. Kimi API price as AA lists it. API-only today; Moonshot says full weights ship by 4 Nov 2026 — move to open weight when they land.	Moonshot / Kimi	Frontier closed	46	1M	$2	$10	—	23 Oct 2026
Muse Wonder Meta's new API flagship for agents and computer use; replaces Muse Spark as Meta's top paid model. Scored: max muse-wonder-1.0, Standard mode. Cached input $0.20. No Contributor-mode discount.	Meta	Frontier closed	43	1M	$4	$20	—	12 Oct 2026
Muse Spark 1.3 Meta's cheaper API line, now below Muse Wonder. Scored: max Standard mode (not used for training). Cached input $0.15 per 1M. Contributor mode is $0.10 / $0.20.	Meta	Frontier closed	36	1M	$1.25	$4.25	—	2 Sep 2026
Claude Haiku 5.5 Anthropic's small model for high-volume, latency-sensitive work; first Haiku with an effort setting. Scored: max effort List for prompts up to 100k tokens; prompts over 100k bill $0.50 / $2.50 per 1M. Cache reads $0.01 / $0.05 per 1M. 128K max output.	Anthropic	Frontier closed	36	1M	$0.10	$0.50	—	7 Oct 2026
Grok 4.7 xAI's (now SpaceXAI) current coding and agent flagship. Scored: high (default); xhigh scores 34 Under 200k prompt tokens ($0.50 cached); $4 / $12 above. Grok 5 was previewed 14 Oct 2026 but is not in the API.	xAI	Frontier closed	35	500k	$2	$6	—	21 Sep 2026
Qwen3.8 Max (0902) Alibaba's proprietary API flagship; the highest-scoring Qwen on Index v5.0. Alibaba Cloud Model Studio international list.	Alibaba / Qwen	Frontier closed	34	984k	$2	$6	—	2 Sep 2026
DeepSeek V4.1 Flash 1017 Top open-weight model on Index v5.0 (tied with GLM-5.3), at a fraction of the price. Scored: max effort First-party API as measured by Artificial Analysis. The deepseek-flash alias routes here.	DeepSeek	Open weight	32	1M	$0.30	$1.20	MIT	17 Oct 2026
GLM-5.3 Zhipu's open-weight frontier; ties DeepSeek for the open-weight lead on v5.0. Scored: max effort	Zhipu / Z.AI	Open weight	32	1M	$1.40	$4.40	GLM-5 license	18 Aug 2026
Gemini 3.8 Flash Google's current Flash; still the cheap, fast Gemini until a Gemini 4 Flash ships. Scored: high Introductory Google list through 31 Dec 2026; then $1.50 / $7.50 from 1 Jan 2027.	Google	Frontier closed	31	1.05M	$0.75	$3.75	—	2 Sep 2026
MiMo-V2.6-Pro Xiaomi's 1.02T MoE (42B active) with native text, image, video, and audio input; third-best open weights on v5.0. API price as AA lists it.	Xiaomi	Open weight	30	1M	$0.43	$0.87	MIT	21 Sep 2026
MiniMax-M3.1-Flash 428B MoE (23B active) for coding and agents; near the open-weight top for pennies. Cached input $0.015 per 1M.	MiniMax	Open weight	29	1M	$0.15	$0.60	MiniMax Community License	19 Oct 2026
GPT-6 Luna OpenAI's volume tier; replaces GPT-5.6 Luna at half the price. Scored: max effort Cached input $0.01 per 1M. Prompts over 272k input tokens bill 2× input.	OpenAI	Frontier closed	28	1.05M	$0.10	$0.50	—	22 Sep 2026
Mistral Large 4 "Le Chonk": Mistral's 1.05T MoE (52B active) flagship, now open weight; replaces Medium 3.5 here. Scored: high Cached input $0.07 per 1M. API public preview from 6 Oct 2026; weights released 29 Oct 2026. Commercial model-as-a-service use needs a separate Mistral license.	Mistral	Open weight	26	1M	$0.70	$2.10	Mistral Ecosystem Open-Weight License	29 Oct 2026
Muse Spark Open Meta's downloadable model now (816B MoE, 42B active); replaces Llama 4 as the Meta open-weight row. Scored: xhigh Hosted-provider price as AA lists it; not on Meta Model API. Weights on Hugging Face.	Meta	Open weight	25	1M	$1.25	$4.25	Muse Model License	15 Oct 2026
DeepSeek V4 Pro 0813 1.6T MoE, 49B active, MIT weights; DeepSeek's big model, now behind its own Flash on v5.0. Scored: max effort First-party API as measured by Artificial Analysis. The newer V4.1 Flash 1017 now outscores it.	DeepSeek	Open weight	23	1M	$1.32	$3.96	MIT	13 Aug 2026
Kimi K3 Moonshot's current downloadable model (2.8T MoE, 104B active). Not yet scored on AA Index v5.0. Kept until Kimi K3.1 Pro weights ship (promised by 4 Nov 2026).	Moonshot / Kimi	Open weight	—	1.05M	$3	$15	Kimi K3 License (commercial with restrictions)	16 Jul 2026
Qwen3.8 2.4T A95B Alibaba's largest open-weight Qwen (2.4T MoE, 95B active). Not yet scored on AA Index v5.0. AA now lists a 262k context for this model.	Alibaba / Qwen	Open weight	—	262k	$2	$6	Qwen3 license	12 Aug 2026
Gemma 4 31B Google's current Gemma dense open-weight; Apache 2.0, 256k context. Not yet scored on AA Index v5.0. Self-host; no Google list price.	Google	Open weight	—	256k	—	—	Apache 2.0	2 Apr 2026

Sources

Figures as of 29 Oct 2026. They lag the labs. We fetched the pages below; we did not invent scores, windows, licenses, or list prices. Every Index score is Artificial Analysis v5.0. Gemini 4 Argon, Gemini 4 Carbon, and Gemini 3.8 Flash show Google's introductory list prices; Artificial Analysis lists the post-introductory price for the Gemini 4 models ($4 / $20).

Artificial Analysis LLM Leaderboard — Intelligence Index scores, context windows, and AA-measured API prices. Every score on this board is Index v5.0 (fetched 29 Oct 2026). v5.0 is a new, harder composite — scores are much lower than the v4.x numbers we showed before and are not comparable to them.
Artificial Analysis Intelligence Index v5.0 — Composite of AA-Briefcase v1.1, AutomationBench-AA, Terminal-Bench 4.0, FrontierCode 2.0, AA-Omniscience v1.1, GDP.pdf, HLE-Diamond, CritPt v2 (Hard Subset), and Terminal-Bench-Science 0.1. Knowledge work, coding, science, and general each weigh 25%.
Google — Introducing Gemini 4 Carbon — Announced 28 Oct 2026. #1 on AA Index v5.0 at 59. Limited to cyber defenders via the Fairwind Program for now. Introductory $2 / $10 per 1M (same as Argon); $4 / $20 after the introductory period.
Google — Gemini 4 Argon is now available for everyone — Generally available 13 Oct 2026. Introductory $2 / $10 per 1M through 30 Nov 2026, then $4 / $20. Cached input 95% off. 1M unified input-output limit.
Anthropic — Introducing Claude Opus 5.5 — Released 22 Sep 2026. $4 / $20 per 1M (20% below Opus 5); cache reads $0.20. 1M context, 128K max output.
Anthropic — Introducing Claude Sonnet 5.5 — Released 28 Sep 2026. $2 / $10 per 1M; cache reads $0.10. 1M context.
Anthropic — API pricing — Current list: Fable 5.1 / Mythos 5.1 $10 / $50, Opus 5.5 $4 / $20, Sonnet 5.5 $2 / $10, Haiku 5.5 $0.10 / $0.50 (≤100k prompt) and $0.50 / $2.50 (>100k).
Anthropic — Claude Fable 5.1 and Mythos 5.1 — Announced 1 Sep 2026. Identical weights; Mythos is limited-availability with a more permissive safeguard layer.
Anthropic — Introducing Claude Haiku 5.5 — Announced 7 Oct 2026. 1M context; 128K max output.
OpenAI — GPT-6 Astra — 1,050,000 context. $10 / $50 per 1M (short context); long context $20 / $75.
OpenAI — Introducing GPT-6.1 Sol — Released 29 Sep 2026. $2 / $10 per 1M; cached input $0.10. 1,050,000 context per OpenAI model docs.
OpenAI — Introducing GPT-6.1 Terra — Released 26 Oct 2026. $1 / $5 per 1M (model docs); cached input $0.05. 1,050,000 context.
OpenAI — Introducing GPT-6 Sol and Luna — Released 22 Sep 2026. GPT-6 Luna is $0.10 / $0.50 per 1M; 1,050,000 context. GPT-6 Sol was superseded a week later by GPT-6.1 Sol.
Meta — Muse Wonder and Muse Wonder Full — Released 12 Oct 2026. muse-wonder-1.0 on Meta Model API at $4 / $20 per 1M (cached $0.20), 1M context. Muse Wonder Full is Muse Code subscription-only.
Meta — Muse Spark Open — Open weights released 15 Oct 2026. 816B MoE / 42B active, 1M context. Served via hosted providers, not Meta Model API.
Meta Model API — pricing — Muse Spark 1.3 Standard mode $1.25 / $4.25 per 1M; cached $0.15.
Kimi — Introducing Kimi K3.1 Pro — Released 23 Oct 2026 on the Kimi API. Post-trained K3 (3T class), 1M context. Full weights promised by 4 Nov 2026.
Mistral — Mistral Large 4 — 1.05T MoE / 52B active, 1M context. API $0.70 / $2.10 per 1M. Open weights (Mistral Ecosystem Open-Weight License) released 29 Oct 2026.
DeepSeek — V4.1 Flash 1017 — 17 Oct 2026 update to V4.1 Flash; deepseek-flash alias routes here. MIT weights.
Xiaomi — MiMo-V2.6 — MiMo-V2.6-Pro open weights released 21 Sep 2026; 1.02T MoE / 42B active; MIT; 1M context.
MiniMax — M3.1 Flash — Released 19 Oct 2026. 428B MoE / 23B active; 1M context; $0.15 / $0.60 per 1M.
Alibaba Cloud Model Studio pricing — Qwen3.8-Max (0902 snapshot, 2 Sep 2026) at $2 / $6 per 1M international.
xAI API release notes — grok-4.7 is the current flagship: 500k context, $2 / $0.50 cached / $6 under 200k prompt tokens, $4 / $1 / $12 above. No Grok 5 in the API as of 29 Oct 2026.
Google — Gemini API pricing — Gemini 3.8 Flash introductory $0.75 / $3.75 per 1M through 31 Dec 2026; then $1.50 / $7.50.
Z.ai — GLM-5.3 — Open-weight GLM-5.3; AA lists 18 Aug 2026 and $1.40 / $4.40 per 1M.

Models

Sources

Know what matters in AI.