Stoichiometry of C and H in rumen fermentation is a key-point to understand the partition of degradable dietary compounds between useful fermentation products (volatile fatty acids or VFA and microbial biomass) and waste products such as methane (CH4) and carbon dioxide (CO2). In this context, in vitro batch methods can be useful because VFA and gas productions are much easier to measure compared to in vivo methods. Fermentation of two substrates with very different carbohydrate composition corresponding to two diets with either 35 or 70% of concentrate, with or without oil addition was studied in vitro with the Hohenheimer Futterwert Test (HFT) procedure. Volatile fatty acid production was measured after three hours of incubation, such as gas production and composition (CH4, CO2 and hydrogen or H2). Gas production and composition, and molecular H2 productions were calculated from VFA production according to classic stoichiometric relationships 1, 2. In vitro gas production has two origins: degradation of substrates (direct gas) and buffering of VFA (indirect gas). The prediction of total gas was accurate with a slope equal to 1 when compared with the measured data, which partly validates chosen stoichiometry coefficients. However, the actual CH4 prediction was curvilinearly linked with the measured one. This response, according to decreasing marginal returns, could be explained by other molecular H2 sinks (microbial synthesis, acetogenesis, fatty acid biohydrogenation, ...). Further, chosen stoichiometric coefficients overestimated theoretical CH4 production in comparison with the measured one. This can question this usual way to model CH4 production. This work also shows that indirect gas might be important (51.2 ± 2.60% of total gas production) that could explain some discrepancies in the HFT method when predicting intake or digestibility of feeds.