model pricing

DeepSeek

2026-09-11 05:11:13

DeepSeek V4.1 Flash trails Kimi and GLM in third-party test, but runs at about one-seventh the cost

Artificial Analysis has completed an independent test of DeepSeek V4.1 Flash, giving the model a composite intelligence score of 40. That puts it ahead of DeepSeek’s own V4 Pro 0813, which scored 36, but still below Kimi K3 at 44 and GLM-5.3 at 45. By that third-party ranking, DeepSeek has not reclaimed the top spot among domestic models. The price gap was far wider than the score gap. DeepSeek V4.1 Flash cost an average of $0.27 to complete one Intelligence Index task, while Kimi K3 and GLM-5.3 were both around $2, a difference of about 7x. Output speed also stood out at 197 tokens per second, compared with roughly 36 tokens per second for Kimi K3 and about 58 tokens per second for GLM-5.3. Its strongest area remained agent performance. On AutomationBench-AA, V4.1 Flash scored 69%, matching GPT-6 Astra and beating GLM-5.3’s 62%. It also reached 84% on the long-context AA-LCR test. One drawback noted in the test was verbosity: in AA tasks, the model produced about 89,000 tokens per task on average, 62% more than V4 Pro, which reduced part of its pricing advantage.

810

B.AI

2026-09-02 07:25:00

B.AI to End Free Access to DeepSeek V4 Flash on Sept 3, Start Discounted Billing

According to ChainCatcher, AI model platform B.AI has set a September 3 deadline for ending the free trial of DeepSeek V4 Flash and DeepSeek V4 Flash Vision Exp. Starting at 17:00 SGT that day, the two models leave the free tier and move to a paid plan billed at a discount. Before that timestamp, users can still access both without charge. B.AI said it would adopt DeepSeek's official peak and off-peak pricing and then stack another discount on top. During peak windows, which run from 09:00 to 12:00 and 14:00 to 18:00 SGT on weekdays, usage is charged at 50% of the official price. During off-peak windows, covering remaining weekday hours and the whole weekend, the price drops by another 50%, making the final effective rate 25% of the official price. At the same time, free access stays in place for other models. GLM 5.3 Flash continues to be free, carrying 320 billion total parameters, 18 billion active parameters, and 1M context support. Qwen3.8 Flash, Hy3 and MiMo V2.5 will also be available for free going forward. ChainCatcher first reported the announcement.

890

Google

2026-08-16 16:02:49

Gemini 3.7 Flash review: big coding gains, weaker reasoning and writing still show

Google launched Gemini 3.7 Flash on August 13 and made it generally available in more than 160 countries on day one. According to Decrypt’s review, the model accepts up to 1 million input tokens, returns 64,000 output tokens, handles images, video, audio, and PDFs, and can use tools while operating a computer. Google’s own benchmark sheet says the model beats Claude Sonnet 5 and GPT-5.6 Terra in 11 of 18 tested categories, including 1,588 Elo on Code Arena’s web development board and 30.4% on AutomationBench, though Decrypt notes those figures come from Google’s methodology and should be treated as company claims rather than settled fact. Decrypt’s hands-on tests found the sharpest improvement in coding. Gemini 3.7 Flash generated a playable browser game on the first try in 2 minutes and 13 seconds, a major step up from Gemini 3.6 Flash, which Decrypt said could not produce a working file in a similar test after its July 21 release. Results were less convincing elsewhere. In creative writing, Decrypt said Gemini produced a tidy story but broke the central prompt rule, losing to a free community model, Qwopus3.5-27B-v3. In associative reasoning, logic, and advanced math, the review said Gemini often showed decent structure but failed on crucial task requirements, including a bridge puzzle and a polynomial problem it left unfinished. Decrypt’s conclusion: Gemini 3.7 Flash is a strong low-cost execution model inside Google’s ecosystem, but its creativity and reasoning remain uneven.

1550

Google

2026-08-14 10:06:51

Google launches Gemini 3.7 Flash three weeks after 3.6, with lower pricing aimed at coding and agent work

Google has released Gemini 3.7 Flash on Aug. 13, shortening its model update cycle to roughly three weeks after Gemini 3.6 Flash. According to Google’s announcement, the new model is positioned as a high-value offering for coding and agent tasks, while also targeting document-heavy knowledge work and web development. It supports text, image, audio, and video input, comes with a 1 million-token context window, and can generate up to 64,000 output tokens. Google said Gemini 3.7 Flash improved on several benchmarks versus Gemini 3.6 Flash, including FrontierCode 1.1, which rose from 34.4% to 43.6%, AutomationBench from 17.0% to 30.4%, and the document-understanding benchmark GDP.pdf from 22.0% to 34.0%. The product is being offered through API and enterprise channels, including Gemini API, Google AI Studio, Antigravity, Android Studio, and Gemini Enterprise. Consumer access is available through Gemini Spark under AI Pro and Ultra plans. Google is not releasing open-weight access for the model. Pricing is a central part of the launch. Through Dec. 31, 2026, input costs are set at $0.75 per 1 million tokens and output at $3.75, before rising to $1.5 and $7.5 in 2027. Using an 80/20 input-output mix, ABMedia estimated blended cost at about $1.35 per 1 million tokens, below Sonnet 5 at $3.60 and GPT-5.6 Terra at $4.00.

1400

DeepSeek

2026-08-13 03:13:21

DeepSeek rolls out V4 Pro official release, claiming near-Claude performance at roughly 1/46 the cost

DeepSeek has swapped in the "0813" official release of deepseek-v4-pro, replacing the preview version that had been running since April. Citing a report from Decrypt, the company’s own benchmark table frames the pitch in simple terms: the model trails Anthropic’s Claude Fable 5 by about 5.3% on average across nine agent benchmarks, while costing only a fraction as much. DeepSeek said V4 Pro supports a 1 million-token context window, and in two of the nine tests it scored above Claude Fable 5. The larger gap is in pricing. DeepSeek lists input at $0.43 per million tokens and output at $0.87 per million tokens, versus $10 input and $50 output for Claude Fable 5. On that basis, the report says DeepSeek’s cost comes out to about 1/46 of Claude Fable 5. DeepSeek also said the official version posted marked gains over the preview build in coding and agent-related benchmarks, including Terminal Bench 2.1, DeepSWE, and DSBench-FullStack. The comparison, however, comes from DeepSeek’s own materials and has yet to be verified by independent third parties.

1510

Alibaba

2026-08-04 01:51:01

Alibaba unveils Qwen3.8-Max with self-reported benchmark lead and aggressive token pricing

Alibaba’s Qwen team has introduced Qwen3.8-Max, a new flagship model that the company says scored 86.1 on OSWorld-Verified, ahead of GPT-5.6 Sol Max at 83.2, Fable 5 at 85.0, and Gemini 3.1 Pro at 76.2. The release also included a broader slate of benchmark claims, such as 93.0 on PaperBench and 86.6 on TerminalBench 2.1, alongside positioning the model for long-running autonomous work rather than standard chatbot use. The model uses a mixture-of-experts architecture with 2.4 trillion total parameters and about 95 billion active during inference, built on the Qwen3.5 architecture with a 1 million-token context window. Alibaba also said Qwen3.8-Max is suited for extended coding tasks, desktop software operation, experiment reproduction, and industrial workflows that feed visual input back into a planning loop. Pricing appears to be a central part of the launch. According to QwenCloud pricing cited in the report, Qwen3.8-Max costs $2 per million input tokens and $6 per million output tokens overseas, bringing the combined total to $8 per million tokens. That is below one-third of Claude Opus 5’s combined $30 and below one-quarter of GPT-5.6 Sol standard mode at $35. Still, the benchmarks and capability demonstrations were all disclosed by Alibaba and have not been independently verified, while the company has yet to publish the licensing terms for the promised open-weight release next week on Hugging Face and ModelScope.

1970

OpenAI

2026-07-31 01:24:52

OpenAI cuts prices for parts of the GPT-5.6 lineup and rolls out a new API mode

OpenAI said it is adjusting pricing and performance across parts of its GPT-5.6 model family in a move aimed at improving cost efficiency for enterprise AI use cases. According to the company’s post on X, the price of GPT-5.6 Luna has been reduced by 80%, while GPT-5.6 Terra is now 20% cheaper. OpenAI also said the lower pricing will be reflected in how paid subscription usage is counted for Codex and ChatGPT Work. Separately, the company introduced a Fast mode in its API, replacing the earlier Priority Processing service. Under Fast mode, GPT-5.6 Sol can process requests at up to 2.5 times the speed of the standard mode. That option is priced at twice the standard processing rate, while the model’s capabilities remain unchanged.

1960