Baidu has rolled out two new models, ERNIE X1 Turbo and ERNIE 4.5 Turbo, boasting impressive performance enhancements alongside substantial cost reductions.
These Turbo versions, crafted as faster and smarter upgrades to the earlier ERNIE X1 and 4.5 models, spotlight improvements in multimodal processing and reasoning skills, all while embracing aggressive pricing meant to woo developers and expand market presence.
The ERNIE X1 Turbo is positioned as a deep-thinking, high-reasoning model, tailored for intricate tasks demanding sophisticated comprehension. In a crowded and competitive landscape, it claims a lead over rivals like DeepSeek R1, V3, and OpenAI’s o1 on several benchmark tests.
At the heart of X1 Turbo’s upgraded capabilities lies a more advanced "chain of thought" reasoning method, enabling it to tackle problems in a structured and logical way.
Alongside this, the model demonstrates beefed-up multimodal capabilities—going beyond text to potentially process images and other types of data—with sharpened skills in tool usage. These enhancements make it an ideal fit for complex applications such as literary creation, intricate logical reasoning, code writing, and nuanced instruction following.
All this capability comes at a surprisingly low cost. Pricing starts at $0.14 per million input tokens and $0.55 per million output tokens—roughly a quarter of what DeepSeek R1 charges.
Sharing the stage, ERNIE 4.5 Turbo brings its own set of upgrades, centering on stronger multimodal abilities and significantly faster response times than its non-Turbo predecessor. Here, Baidu leans heavily into creating a versatile and nimble AI experience while dramatically slashing costs.
The 4.5 Turbo model cuts pricing by a staggering 80% compared to the original ERNIE 4.5, setting input token rates at $0.11 per million and output at $0.44 per million—about 40% of DeepSeek V3’s latest rates.
Performance tests back up these claims. In evaluations measuring both text and multimodal skills, ERNIE 4.5 Turbo edges out OpenAI’s acclaimed GPT-4o model. In particular, for multimodal tasks, ERNIE 4.5 Turbo posted an average score of 77.68, surpassing GPT-4o’s 72.76.
Of course, while benchmarks should always be interpreted with caution, these results position ERNIE 4.5 Turbo as a serious contender for tasks requiring integrated understanding across multiple data formats.
The debut of ERNIE X1 Turbo and 4.5 Turbo underlines a wider shift in the AI world: a move toward democratizing cutting-edge capabilities. As foundational models continue pushing technical boundaries, the appetite for affordable yet powerful AI solutions is only growing.
By slashing prices without compromising on sophistication, Baidu’s ERNIE Turbo series may well open the doors for a broader set of developers and companies to embrace advanced AI.
This pricing strategy also ramps up the pressure on established giants like OpenAI and Anthropic, as well as emerging challengers like DeepSeek, potentially triggering a fresh wave of market-wide price cuts.
